Image compression

The Single Pass Recompression (SPR) method optimizes image compression in raster image processing systems by classifying areas for lossless and lossy compression, addressing inconsistent quality issues and achieving target storage sizes efficiently in a single pass.

DE102013203884B4Active Publication Date: 2024-11-14CANON KK
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
DE102013203884
Authority / Receiving Office
DE · DE
Patent Type
Patents
Current Assignee / Owner
Priority Date
2012-03-21
Filing Date
2013-03-07
Publication Date
2024-11-14
Estimated Expiration
2033-03-07

AI Technical Summary

Technical Problem

Existing image compression methods in raster image processing systems face challenges in predicting memory savings during recompression, leading to inconsistent image quality and the need for multiple passes to achieve target storage sizes, especially when dealing with diverse image types like text, graphics, and natural images.

Method used

The Single Pass Recompression (SPR) method determines appropriate recompression parameters to ensure that the compressed data meets storage requirements while optimizing image quality by classifying areas as lossless or lossy and adjusting compression parameters in a single pass, using edge-based lossless compression for text and graphics and lossy compression for natural images.

Benefits of technology

The SPR method effectively achieves target storage sizes with minimal quality loss by strategically managing lossless and lossy compression, ensuring consistent image quality across the page in a single pass, reducing computational intensity and memory usage.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 00000000_0000_ABST
    Figure 00000000_0000_ABST
Patent Text Reader

Abstract

A method for compressing an image to be stored in memory in order to satisfy a memory requirement, wherein the method comprises the following steps: Determining the size of an area in the image, where the area has a uniform color (1201), Obtaining value candidates from a range size threshold (1300, 1390, ES_Threshold), characterized by Comparing the determined size of the range with at least one of the value candidates of the range size threshold (1202), Estimating data sets (1901, 1902, 1904) for encoding edges that define the range satisfying at least one of the candidate values ​​of the range size threshold, Selecting a value from the candidate values ​​as the range size threshold based on the comparison between the estimated data amounts and the storage requirement (1903), lossless compression of the area-defining edges that meets the selected value of the area size threshold (1805), and Lossy compression of image data in a region excluding the region to be compressed without loss (1805).
Need to check novelty before this filing date? Find Prior Art

Description

Reference to related patent applications

[0001] This application claims the benefit under 35 USC §119 of the filing date of Australian Patent Application No. 2012201684, filed on March 21, 2012, which is hereby incorporated by reference in its entirety as if fully set forth herein. Technical field

[0002] The present invention relates to image processing and, in particular, to a method and a device for compressing images. The present invention also relates to a computer program, including a computer-readable medium containing a computer program for compressing images. background

[0003] Many computer systems are needed to process large amounts of image or dot-matrix data (bitmap data). Such computer systems often operate with limited memory resources, and therefore these computer systems often compress data to reduce the amount of memory required during processing.

[0004] An example of such a computer system is a raster image processing system. A raster image processor (RIP) takes the digital information regarding fonts and graphics that describe the appearance of a document and translates this information into an image consisting of individual pixels (picture elements) that the image-generating device, such as a printer, can output. Several types of raster image processors are known in the art. These include frame-memory RIPs, tape-memory RIPs, and tile-sequence RIPs.

[0005] In a tile-based RIP, the page is divided into square tiles. Each tile is fully rendered before the RIP begins rendering the next tile. With such a RIP, an output compression device can begin compressing a rendered tile as soon as it is available, provided the compression device can accept data in tile order.

[0006] When selecting a compression method, it is important to use one that is suitable for the type of data being compressed. These methods can generally be categorized as either lossy or lossless.

[0007] A hybrid compression strategy, employing both lossless and lossy compression algorithms, is commonly used to compress the output of a raster image processor where the rendered dot graphics are segmented into two-dimensional and two-dimensional areas. In principle, this is a successful strategy for achieving high image quality while maintaining a reasonable compression ratio. However, in a system where memory resources are limited, adjustments to the compression strategy are necessary to achieve compression within a target memory size (or permissible memory capacity).

[0008] For example, in a typical JPEG compression process, the DCT coefficients are quantized using a quantization table, which can be used to control the size of a final encoded image. The higher the quantization level, the more noticeable the loss of information during quantization.

[0009] A hybrid strategy using JPEG's lossy compression algorithm can adjust its lossy compression ratio by modifying the quantization level. If the initial compression fails to meet the target memory size, recompression with a higher quantization level is required. However, predicting the memory savings achievable by setting a specific quantization level is difficult. Therefore, multiple recompression attempts are needed, each incrementally adjusting the quantization level until a level suitable for a given target memory size is reached.

[0010] Another method for controlling the size of an encoded image is to discard the image's DCT coefficients, starting with the most significant coefficient and stopping when a desired target memory size is reached. This is possible because the majority of the visual information in the image is encoded within the DC and lower AC coefficients.

[0011] A method of spectral JPEG encoding of images can be used in which higher-order AC coefficients can be deleted without the need for recompression. For each DCT encoding block, the quantized coefficients are stored within spectral bands of decreasing visual significance. The ascending order of the DCT coefficients is directly related to their decreasing influence on the visual quality of the image. Therefore, the encoded DCT coefficients can be grouped into spectral parts sorted according to decreasing relevance to the overall image quality. With this arrangement, image quality can be reduced as little as possible by deleting the higher-order spectral parts and their associated encoded DCT coefficients. This makes it possible to reclaim memory appropriately by first removing the less relevant parts, and there is no need for recompression.Furthermore, in order to reduce the size of the losslessly compressed data, the processing unit segments the image by setting a more aggressive area segmentation parameter, so that more areas than image areas can be classified in the recompression process.

[0012] A method for classifying more areas than image areas dynamically adjusts both compression and segmentation parameters to maintain the instantaneous compression ratio within a tolerance range. The method compresses the document strip by strip until the overall compression ratio meets the requirement. However, adjusting compression and segmentation parameters strip by strip can lead to inconsistent quality across the page. Recompressing based on an instantaneous compression ratio rather than the final overall compression ratio can lead to premature optimization and unnecessary quality loss. In most cases, the entire image can potentially be recompressed more than once, as it is difficult to determine the memory savings that can be achieved by adjusting such a parameter before recompression occurs.

[0013] Another method discloses a block-based hybrid compression system where the compression ratio is controlled by adjusting the threshold used to classify blocks as persistent or as images. A mapping between the number of stable blocks and a value of the classification threshold is stored. Each point in the mapping indicates the number of blocks that can be classified as persistent when such a classification threshold is selected. This information provides a rough indication of the compression ratio for a given classification threshold, which is then used to adjust the threshold when recompression is needed. However, a compression ratio cannot be accurately predicted simply by using the number of stable blocks, meaning that an attempted recompression may fail to achieve the desired compression ratio.Other methods for reducing memory usage when recompression fails to achieve the desired compression ratio include image degradation, increased quantization step size, or selective quantization for darker areas. The disadvantage of these other methods is the need for additional recompression passes, which must be performed until the desired compression ratio is achieved.

[0014] From US patent 2008 / 0144953A1, an image processing method and an image processing device for processing a compressed image generated by segmenting image data into a plurality of MCUs (Minimal Coded Units) and using a compression method in which a frequency transformation is applied to each MCU is disclosed. When the compressed image is decoded, each MCU is classified according to the frequency coefficients it contains, and color-separated data is generated for each MCU based on the decoded data resulting from the decoding and the results of the MCU classification.

[0015] Furthermore, US Patent 6,373,583 B1 discloses a data compression technique for composite document pages. Specifically, it discloses a fast, raster-based, enhanced data compression technique for printing composite documents, including preprocessing images in an original page description form of the data prior to rasterizing the page data. This significantly improves compressibility. Unscaled image data is filtered before being rasterized to the final print resolution. This filtering allows a separate, near-lossless compression algorithm to be applied to a rasterized page description at high compression rates. This data compression technique enables a system to compress composite documents using a relatively fast and simple algorithm, while maintaining near-lossless print quality.

[0016] Furthermore, US Patent 2003 / 0081845A1 discloses a single-pass data compression method with guaranteed fit using rate feedback. In this method, an image compression system provides high-quality, single-pass, block-by-block compression with guaranteed fit. An output block target assigner determines an output target size for the compressed blocks, derived from the source image blocks. A block target size summer determines a target block size by adding any excess compression from the preceding block to the output, as indicated by a compressed block size reader. A mode selector performs a content analysis of the current block. The mode selector then chooses from among the compression modes that are guaranteed to meet the target block size, selecting one that achieves optimal image quality.A compression encoder compresses the current block according to the selected compression mode. Summary

[0017] It is an object of the present invention to substantially overcome or at least improve one or more disadvantages of existing arrangements.

[0018] Arrangements referred to as single-pass recompression (SPR) arrangements have been revealed, which attempt to address the above problems by determining a target compression size when a storage capacity (or target storage size) is exceeded during an initial compression and, depending on this, changing the mixture of edge data compression and image data compression methods in order to perform only one further compression to achieve a utilization of the target storage.

[0019] According to one aspect of the invention, a method for compressing an image to be stored in memory is provided in order to meet a storage requirement, wherein the method comprises the following steps: Determining the size of an area in the image, where the area has a uniform color, Obtaining value candidates within a range size threshold, Comparing the determined size of the range with at least one of the value candidates of the range size threshold, Estimating data sets for encoding edges that define the range satisfying at least one of the candidate values ​​of the range size threshold, Selecting a value from the candidate values ​​as the range size threshold based on a comparison between the estimated data volumes and the storage requirement, lossless compression of the area-defining edges that meet the selected value of the area size threshold, and Lossy compression of image data in an area excluding the area to be compressed losslessly. Brief description of the drawings

[0020] One or more embodiments of the invention will now be described with reference to the following drawings, in which: Fig. 1 A schematic block diagram of a printing system for rendering and printing a document shows, Fig. 2 a schematic block diagram of the printing system Fig. 1 with a client-based architecture shows, Fig. 3 a schematic block diagram of the printing system Fig. 1 with a host-based architecture shows, Fig. 4 depicts a tile, Fig. 5 A schematic block diagram for a RIP output compression device as used in the systems from Fig. 2 and Fig. 3 is used, according to an SPR arrangement, Fig. 6 a schematic flowchart of a process through which Fig. The RIP output compression unit shown in Figure 5 demonstrates the compression process in progress. Fig. 7. An extended view of the schematic block diagram for a [unclear] Fig. Figure 5 shows a RIP output compression device according to an SPR arrangement, Fig. Figure 8 shows a flowchart of a pixel array generation process for a single pixel array on a sample row (or column) of a tile, Fig. 9 shows the neighborhood of pixels that are used to apply contrast to a row of pixels, Fig. 10 examples of edge formation processes are shown in which the merge criteria are met by the pixel sequence candidates. Fig. 11 A flowchart of an edge generation process performed by the edge generation device shows, Fig. Figure 12 shows a schematic flowchart of a procedure for assessing the visual significance of edges generated by the edge generation device, Fig. 13a is an example of a representation showing how the edge metadata is stored in the recompression management facility, Fig. 13b an alternative representation for the same in Fig. The example shown in 13a illustrates how the edge metadata is stored in the recompression management facility. Fig. 14A is a schematic diagram of an image subdivided into coding units for lossy encoding, Fig. 14B is a schematic diagram of an 8 by 8 DCT block, Fig. 15 is a schematic diagram showing the arrangement of a lossy image compression device according to an SPR arrangement, Fig. 16a is a schematic diagram of the quality partitions in the DCT domain, Fig. 16b is a table showing an exemplary partitioning scheme for the coding DCT coefficients, Fig. Figure 17 shows an example tile compressed by an SPR arrangement, Fig. 18 is a schematic flowchart that illustrates the in Fig. The strategic storage reduction step shown in Figure 6 illustrates Fig. 19 An exemplary histogram shows that the metadata used to adjust the edge evaluation criteria represents, Fig. 20a is a schematic diagram of a procedure for storing the quality partitions within a memory after an initial compression, Fig. 20b is a schematic diagram of a procedure for storing the quality partitions within a memory after recompression, and Fig. 21A and Fig. 21B form a schematic block diagram of a general computer system in which the described arrangements can be operated. Detailed description including best practice

[0021] Where one or more of the attached drawings refer to steps and / or features that have the same reference numerals, for the purposes of this description these steps and / or features have the same function(s) or the same sequence(s), unless otherwise indicated.

[0022] It is noted that the discussions concerning prior art arrangements contained in the "Background" section and the section above relate to discussions of arrangements that could constitute public knowledge due to their respective uses. Such discussions should not be interpreted as a statement by the inventors or the current patent applicant that such arrangements are in any way part of the common general knowledge.

[0023] Since data recompression is computationally intensive, it is advantageous to determine suitable recompression parameters that guarantee the resulting compressed data meets the storage requirements (allowable storage capacity) in a single pass, while optimizing image quality. The disclosed SPR arrangement (SPR: Single Pass Recompression) solves the problem of how to determine such recompression parameters in a hybrid compression system.

[0024] The principles of the arrangements described herein have general applicability to image compression. For simplicity, the arrangements are described using the image compression method employed in a color raster image processing system. However, the SPR arrangements are not intended to be limited to the arrangements described. For example, the SPR arrangement can be applied in any compression-initiating arrangement where memory resources are limited.

[0025] For natural images, lossy compression algorithms like JPEG and wavelet compression can achieve high compression ratios and acceptable visual quality by discarding information that is not visually significant. However, documents containing sharp color transitions, such as text or graphics, can suffer from the introduction of visible artifacts if information is discarded.

[0026] The advantage of lossless compression is that the output is of high quality. This is important for text and graphics where sharp color transitions must be maintained, and it avoids the types of artifacts caused by most lossy algorithms. Pixel-based lossless methods like JPEG-LS do not scale well, as increasing resolution and color depth makes them extremely memory-intensive. In the worst-case scenario, the compressed file size can be larger than the original file size.

[0027] Edge-based lossless algorithms locate areas of connected pixels and encode the shape of each area. They can render monochrome text or graphic areas very efficiently because a large area containing many pixels can be described with just a single edge. They are less affected by increasing resolution or bit depth because the number of edges does not increase with higher resolution. However, natural images do not compress well with edge-based algorithms.

[0028] A lossless compression method alone produces a satisfactory result for compressing a RIP output. A typical RIP output contains a wide variety of different requirements across a single page. Pixel data generated by a RIP can belong to a number of different area types, such as text, graphics, and natural image areas. Each area type has different characteristics and therefore different compression requirements. A combination or hybrid of lossless and lossy methods is one way to achieve high compression ratios while maintaining high quality, with the lossless method preserving sharp color transitions, while the lossy method provides strong compression for areas with many colors.

[0029] The mixed-form approach requires several methods for identifying which areas should be coded losslessly and which should be coded lossily.

[0030] Areas of pixels that form characters, symbols, and other glyphs are called text areas. Areas of pixels that form large areas of the same color, as commonly found in many block diagrams, charts, and graphics (clip art), are called graphic areas. Both text and graphic areas contain a single color per area and require well-defined transitions between areas to maintain sharp edges. Lossless encoding is typically used for these areas. Text and graphic areas are collectively referred to as "planar areas" and, in this description, also as "first areas."

[0031] Image areas, also referred to as "secondary areas" in this description, are regions containing many colors that transition more smoothly than the color transitions in graphic or text areas. These areas, typically associated with photographic images, contain a large amount of data due to the constantly changing colors within them. Lossy encoding is commonly used for image areas. The boundary of an image area must still be precisely maintained, as the human visual system will perceive the boundary between an image area and a two-dimensional area almost in the same way as the boundary between two two-dimensional areas. However, the pixels within an image area do not need to be precisely maintained, as the human visual system is less sensitive to small variations in color or luminance within an image area.

[0032] Fig. Figure 1 is a schematic block diagram of a typical printing system 100, which includes a personal computer (personal computer, workstation) 101 connected to a printer 105 via a network 104. As shown below with reference to Fig. 21A and Fig. As described in more detail in Section 21B, the personal computer 101 and the printer 105 each contain at least one processor unit, one memory unit, and a modulator-demodulator transmit-receive device (modem transceiver) for transmitting to and from the network 104. The printer 105 further contains a printer module 107. The network 104 can be any connection, including a universal serial port (USB) or a parallel port connection.

[0033] If a user 2168 (see Fig. When the user selects option 21A) of Personal Computer 101 to print a document 2171 onto a physical medium using Printer 105, there are several stages in the process. First, a software application 102 running on Personal Computer 101 generates data in the form of a Page Description Language (PDL) such as Adobe™ PostScript™ or Hewlett-Packard's Printer Command Language (PCL), which describes the objects to be printed. Second, a host print management device 103 (host: service-providing device), also running on Personal Computer 103, processes the PDL before transferring the resulting data from Personal Computer 101 to Printer 105 via the network 104.

[0034] A client print management unit 106 (client: service-requesting device) in the printer 105 performs further processing before the resulting data is made available to the printer module 107 of the printer 105, where the resulting data is printed onto a physical medium.

[0035] The work performed by the host print management unit 103 and the client print management unit 106 typically consists of job creation, raster image processing (RIP), RIP output compression, RIP output decompression, and post-RIP processing. These tasks can be divided between the host print management unit 103 and the client print management unit 106 in a number of different ways, depending on the type of architecture chosen.

[0036] The RIP is responsible for combining the many layers and objects that can exist in a typical print job into a two-dimensional rasterized output. The output must be able to define the color value for each pixel of the page at the selected resolution. Due to the real-time requirements of a laser printer module, the entire page is usually available in raster form as soon as processing from the RIP begins.

[0037] Post-RIP processing is the process of taking the rendered data, performing any necessary final processing, and feeding the data into the printer module 107 in real time. If all stages of the printing process could guarantee a real-time data feed, then a simple one-pass system could operate, with data being routed through the system at a constant rate in time for printing. However, raster image processors do not always operate in real time due to the varying complexity of the source data that needs to be rendered.

[0038] In a typical laser printer module, processing after a RIP must operate in real time, as soon as the page is fed by printer 105. Otherwise, printer module 107 will be delayed, and the entire page will have to be reprinted. To guarantee real-time data input, an entire page of RIP output must be buffered. The memory required to buffer an entire page of uncompressed pixel data is extremely expensive. Therefore, RIP output compression is necessary to achieve adequate performance at a low cost. Decompression of the RIP output must also be performed in real time.

[0039] Delegating tasks to either the host print management facility (103) or the client print management facility (106) depends on the chosen architecture. The two most common architectures are client-based and host-based.

[0040] Fig. Figure 2 shows a schematic block diagram of a client-based architecture for the printing system. Fig. 1, with the majority of the processing being performed by the client print management unit 106. The user 2168 of the personal computer 101 selects to print document 2171, which causes the software application 102 to generate a PDL that is sent to the host print management unit 103. A job generation unit 201 within the host print management unit 103 takes the PDL and arranges it in a format that can be fed to a RIP 203 in the client print management unit 106. From the job generation unit 201, the data is sent over the network 104 to the printer 105, which stores the data in a job memory 202 of the client print management unit 106. The data is then rendered by the RIP 203 to produce a dot graphic made up of pixels, called the RIP output.The RIP output is then compressed by a RIP output compression device 204 and stored in compressed memory 205 (where "compressed memory" refers to a memory area allocated for storing compressed data). Before the printer module 107 needs the information, the data is decompressed by a RIP output decompression device 206 into uncompressed memory 207 (where "uncompressed memory" refers to a memory area allocated for storing uncompressed data). This data is modified by the processor for processing after RIP 208 in a number of ways to optimize the print quality produced by the printer module 107. Finally, the pixel data is fed to the printer module 107.

[0041] Fig. Figure 3 shows a schematic block diagram of a host-based architecture of the printing system. Fig. 1, with a large portion of the processing being offloaded to the host print management unit 103. User 2168 of personal computer 101 selects to print document 2171, which causes software application 102 to generate a PDL that is sent to the host print management unit 103. Job generation unit 201 processes the PDL and arranges it in a format that can be fed to the RIP 203. This data is stored in job memory 202 before being rendered by the RIP 203 to produce a dot graphic of pixels called the RIP output. The RIP output is compressed by the RIP output compression unit 204 and sent over network 104 to the client print management unit 106 in the printer to be stored in compressed memory 205.Before the printer module 107 requires the information, the data is decompressed by the RIP output decompression unit 206 and stored in the uncompressed memory 207. From there, the processor for processing to RIP 208 modifies the data in a number of ways to optimize the print quality produced by the printer module 107. Finally, the pixel data is sent to the printer module 107.

[0042] Fig. 21A and Fig. 21B models represent a universal computer system 2100, on which the various SPR arrangements described can be applied. The with Fig. 21A and Fig. The related description in Section 21B is mainly directed to the structure and function of the Personal Computer 101, but the structure and function of the Printer 105 is quite similar, and accordingly, descriptions of how the Personal Computer operates can be applied to a large extent to the Printer.

[0043] As in Fig. Figure 21A shows the computer system 2100: a computer module 101, input devices such as a keyboard 2102, a mouse pointer 2103, a scanner 2126, a camera 2127 and a microphone 2180, and output devices including the printer 105, a display device 2114 and loudspeakers 2117. An external modulator-demodulator transceiver (modem) 2116 can be used by the computer module 101 for transmission to and from a communications network 104 via a connection 2121.

[0044] Communications network 104 can be a wide-area network (WAN) such as the internet, a cellular telecommunications network, or a private WAN. If connection 2121 is a telephone line, modem 2116 can be a traditional dial-up modem. Alternatively, if connection 2121 is a high-capacity connection (for example, cable), modem 2116 can be a broadband modem. A wireless modem can also be used for a wireless connection to communications network 104.

[0045] The computer module 101 typically contains at least one processor unit 2105 and one memory unit 2106. For example, the memory unit 2106 can include semiconductor read / write memory (random access memory, RAM) and semiconductor read-only memory (ROM). The computer module also includes a number of input / output (I / O) interfaces, including: an audio-video interface 2107, which connects to the video display device 2114, speakers 2117, and microphone 2180; an I / O interface 2113, which connects to the keyboard 2102, mouse 2103, scanner 2126, camera 2127, and optionally a joystick or other (not shown) human interface device (HID); and an interface 2108 for the external modem 2116 and the printer 105. In some implementations, the modem 2116 may be integrated within the computer module 101, for example, within interface 2108.The computer module 101 also has an interface 2111 for a local network, which allows the computer system 2100 to connect via a connection 2123 to a local communication network 2122 known as a Local Area Network (LAN). As in . Fig. As shown in Figure 21A, the local communication network 2122 can also connect to the large area network 104 via a connection 2124, which typically includes a so-called firewall device or a device with comparable functionality. The interface 2111 for a local network can include an Ethernet™ circuit board, a Bluetooth™ wireless array, or an IEEE 802.11 wireless array; however, numerous other types of interfaces can be used for interface 2111.

[0046] The I / O interfaces 2108 and 2113 can provide one or both serial and parallel connectivity, with the former typically conforming to Universal Serial Bus (USB) standards and including appropriate USB connection devices (not shown). Storage devices 2109 are provided and typically include a hard disk drive (HDD) 2110. Other storage devices, such as a floppy disk drive and a magnetic tape drive (not shown), can also be used. An optical disk drive 2112 is typically provided to act as a non-volatile data source. Portable storage devices, such as optical disks (for example, CD-ROM, DVD, Blu-ray Disc™), USB RAM, portable external hard disk drives, and floppy disk drives, can be used as suitable data sources for the System 2100.

[0047] Components 2105 to 2113 of computer module 101 typically communicate via an interconnected bus 2104 in a manner that results in a conventional operating mode of the computer system 2100 known to those skilled in the art. For example, processor 2105 is connected to the system bus 2104 using a connection 2118. Similarly, memory 2106 and optical disk drive 2112 are connected to the system bus 2104 via connections 2119. Examples of computers on which the described arrangements can be operated include IBM PCs and their compatibles, Sun Sparcstations, Apple Mac™, or similar computer systems.

[0048] The SPR method can be implemented using the 2100 computer system, whereby the processes to be described are Fig. 6 to 8, 11 to 12 and 18 can be implemented as one or more software application programs 2133, 2172, which are executable on the computer system 2100. The SPR procedures can be implemented using either program 2133 running on the computer 101 and / or program 2172 running on the printer 105. In particular, the steps of the SPR procedure are defined by instructions 2131 (see Fig. 21B) is achieved in the software 2133, 2172, which is performed in the computer system 2100. The software instructions 2131 can be formed as one or more code modules, each for performing one or more specific tasks. The software can also be divided into two separate parts, a first part and the corresponding code modules executing the SPR procedures, and a second part and the corresponding code modules managing a user interface between the first part and the user.

[0049] The software can be stored on a computer-readable medium, including, for example, the storage devices described below. The software is loaded into the computer system 2100 from the computer-readable medium and then executed by the computer system 2100. A computer-readable medium containing software or a computer program recorded on it is a computer program product. The use of a computer program product in the computer system 2100 preferably results in an advantageous device for carrying out the SPR process.

[0050] The software 2133, 2172 is typically stored in the HDD 2110 and / or the memory 2106 and / or the memory 2170. The software is loaded into the computer system 2100 from a computer-readable medium and executed by the computer system 2100. Therefore, the software 2133 can, for example, be stored on an optically readable disk storage medium (for example, CD-ROM) 2125, which is read by the optical disk drive 2112. A computer-readable medium with such software or a computer program recorded on it is a computer program product. The use of the computer program product in the computer system 2100 preferably evokes a device for carrying out the SPR method.

[0051] In some examples, the application programs 2133 and 2172 may be provided to the user encoded on one or more CD-ROMs 2125 and read via the corresponding drive 2112, or alternatively, they may be read by the user from the networks 104 or 2122. Furthermore, the software may also be loaded into the computer system 2100 from other computer-readable media. Computer-readable storage media refers to any non-volatile physical storage medium that provides recorded instructions and / or data to the computer system 2100 for execution and / or processing. Examples of such storage media include floppy disks, magnetic tape, CD-ROMs, DVDs, Blu-ray™ Discs, a hard disk drive, a ROM or integrated circuit, USB storage devices, magneto-optical disk storage devices, or a computer-readable card such as a PCMCIA card and the like, regardless of whether such devices are inside or outside the computer module 101.

[0052] Examples of transient or non-material computer-readable transmission media that may also be involved in providing software, application programs, instructions and / or data to Computer Module 101 include radio or infrared transmission channels, as well as a network connection to another computer or network-connected device, and the Internet or intranets, including email transmissions (transmissions of electronic mail) and information recorded on web pages and the like.

[0053] The second part of the application programs 2133, 2172, and the corresponding code modules, as mentioned above, can be executed to implement one or more graphical user interfaces (GUIs) to be rendered or otherwise displayed on the display device 2114. By typically manipulating the keyboard 2102 and the mouse 2103, a user of the computer system 2100 and the application can operate the interface in a customizable manner to provide control commands and / or input to the applications connected to the GUI(s). Other types of customizable user interfaces can also be implemented, such as an audio interface that uses spoken messages output via the loudspeakers 2117 and user voice commands entered via the microphone 2180.

[0054] Fig. Figure 21B is a detailed schematic block diagram of the processor 2105 and a "memory" 2134. The memory represents a logical summary of all memory modules (including the HDD (hard disk drive) 2109 and the semiconductor memory 2016) to which the computer module in Fig. 21A can be accessed. The in Fig. The structure and function shown in Figure 21B may also be applicable, with some modifications, to the operation of the printer 105 and the application program 2172.

[0055] When the computer module 101 is initially powered on, a power-on self-test (POST) program 2150 is executed. The power-on self-test (POST) program 2150 is typically located in a ROM 2149 of the semiconductor memory 2106. Fig. 21A. A software-storing hardware device such as ROM 2149 is sometimes referred to as firmware. The POST program 2150 checks hardware in computer module 101 to ensure proper operation, and typically verifies the processor 2105, memory 2134 (2109, 2106), and a basic input-output system (BIOS) software module 2151, also typically stored in ROM 2149, for correct operation. Once the POST program 2150 has completed successfully, BIOS 2151 enables the hard disk drive 2110 from Fig. 21A. Activating the hard disk drive 2110 causes a bootstrap loader program 2152, resident on the hard disk drive 2110, to be executed by the processor 2105. This loads an operating system 2153 into RAM 2106, whereupon the operating system 2153 begins to run. The operating system 2153 is a system-level application executable by the processor 2105 to perform various high-level functions, including processor management, memory management, device management, storage management, software application interface, and general user interface.

[0056] The 2153 operating system manages memory 2134 (2109, 2106) to ensure that each process or application running on computer module 101 has sufficient memory in which to execute without conflicting with memory allocated to another process. Furthermore, the various types of [unclear] in the 2100 system must [unclear] Fig. 21A The available memory is used appropriately so that each process can run effectively. Accordingly, for total memory 2134, it is not intended to show how specific segments of memory are allocated (unless otherwise specified), but rather to provide a general view of the memory that the computer system can access and how such memory is used.

[0057] As in Fig. As shown in Figure 21B, the processor 2105 contains a number of function modules, including a control unit 2139, an arithmetic logic unit (ALU) 2140, and a local or internal memory 2148, sometimes called cache memory. The cache memory 2148 typically contains a number of memory registers 2144–2146 in a register section. One or more internal bus lines 2141 connect these function modules functionally. The processor 2105 also typically has one or more interfaces 2142 for communicating with external devices via the system bus line 2104 using a connection 2118. The memory 2134 is connected to the bus line 2104 using a connection 2119.

[0058] The application program 2133 contains a sequence of instructions 2131, which may include conditional branching and loop instructions. The program 2133 may also contain data 2132, which is used when the program 2133 is executed. The instructions 2131 and the data 2132 are stored in memory locations 2128, 2129, 2130 and 2135, 2136, 2137, respectively. Depending on the relative size of the instructions 2131 and the memory locations 2128–2130, a particular instruction may be stored in a single memory location, as illustrated by the instruction at memory location 2130. Alternatively, an instruction can be segmented into a number of parts, each of which is stored in a separate memory location, as illustrated by the instruction segments shown at memory locations 2128 and 2129.

[0059] Basically, the processor 2105 is given a set of instructions, which it then executes. The processor 1105 waits for a subsequent input, to which the processor 2105 responds by executing another set of instructions. Each input can be provided by one or more of a number of sources, including data generated by one or more of the input devices 2102, 2103, data received from an external source via one of the networks 104, 2102, data retrieved from one of the storage devices 2106, 2109, or data retrieved from a storage medium 2125 inserted into the corresponding read device 2112, all of which are in Fig. Figure 21A illustrates this. The execution of a set of instructions may, in some cases, result in the output of data. Execution may also include storing data or variables in memory location 2134.

[0060] The disclosed SPR arrangements use input variables 2154, which are stored in memory 2134 at corresponding memory locations 2155, 2156, and 2157. The SPR arrangements generate output variables 2161, which are stored in memory 2134 at corresponding memory locations 2162, 2163, and 2164. Intermediate variables 2158 can be stored at memory locations 2159, 2160, 2166, and 2167.

[0061] Referring to the 2105 processor from Fig. In 21B, registers 2144, 2145, 2146, the arithmetic logic unit (ALU) 2140, and the control unit 2139 work together to perform sequences of micro-operations required to execute "fetch-decode-execute" passes for each instruction in the instruction set that constitutes program 2133. Each fetch-decode-execute pass includes: (a) a fetch operation that fetches or reads an instruction 2131 from a memory location 2128, 2129, 2130, (b) a decoding operation in which the control unit 2139 determines which instructions were fetched, and (c) an execution operation in which the control unit 2139 and / or the ALU 2140 executes the instruction.

[0062] After that, another pass of fetch, decode, and execute can be performed for the next instruction. Similarly, a memory pass can be performed, through which the control unit 2139 stores or writes a value to a memory location 2132.

[0063] Each step or subprocess in the processes of Fig. 6 to 8, 11 to 12 and 18 is connected to one or more segments of program 2133 and / or program 2172 and is carried out by the register section 2144, 2145, 2147, the ALU 2140, and the control unit 2139 in the processor, which work together to perform the fetching, decoding, and execution for each instruction in the instruction set for the said segments of program 2133 when reference is made to program 2133, and by corresponding register sections, ALUs, control units, and processors in the printer 105 when reference is made to program 2172.

[0064] The SPR method can alternatively be implemented in dedicated hardware, such as one or more gate arrays and / or integrated circuits that execute the SPR functions or sub-functions. Such dedicated hardware can also include graphics processors, digital signal processors, or one or more microprocessors and associated memory. When gate arrays are used, the process flow diagrams are shown in Fig. 6 to 8, 11 to 12, and 18 are converted into Hardware Description Language (HDL). This HDL description is then converted into a device-level netlist, which is used by a place and route tool to generate a file that is downloaded to the gate array for programming with the design specified in the HDL description.

[0065] The SPR arrangements include a number of compression algorithms for compressing rasterized data using a combination of lossless and lossy compression, and post-compression memory recovery strategies. In the disclosed examples, the rasterized data is fed to the compression algorithm in a tile-by-tile order. However, the rasterized data can also be provided in other forms, such as stripe-by-stripe.

[0066] The disclosed examples of the SPR arrangement compress images as tiles. For the purposes of this arrangement, a tile refers to a block of N by M pixels, with multiple blocks along the width of the page and multiple blocks along the length of the page. Tiles are separate and cover the page. A tile preferably consists of an integer number of 8x8 blocks of pixels, such as a block of 402. For example, for an A4 page at a printer resolution of 600 dpi, a suitable choice for tile dimensions is M = N = 64. The position of a pixel (X, Y), where X and Y are integers, within a tile is relative to the upper-left corner of the tile. Y indexes the tile rows, with X indexing the offset of a pixel along a tile row. A tile row consists of the set of pixels spanning the width of the tile.

[0067] Fig. Figure 4, for example, shows a tile 400 with 64x64 pixels, where the first pixel 401 in the first tile row occupies a pixel position (0, 0), and the last pixel 403 in the first tile row occupies a pixel position (63, 0). Similarly, the last pixel 404 in the last tile row occupies a position (63, 63). Raster tile order refers to processing a tile pixel by pixel, tile row by tile row, in sequential order, starting with the first tile row and ending with the last row. Pixel values ​​PX, Y within a tile refer to the color value of pixel P located at a position (X, Y). Where the dimensions of a page do not contain an integer number of tiles, the page is preferably padded to the required size. Typically, tiles are processed one after the other, but they can also be processed in parallel.

[0068] Fig. Figure 5 shows an example of a functional block diagram of the RIP output compression device 204, wherein the compression device includes a hybrid compression and decompression device 502, a recompression management device 503 and a compression memory management device 504.

[0069] The RIP output compression unit 204 receives rasterized data 501 from the RIP 203 as input. The data is passed to the hybrid compression / decompression unit 502 for compression. The hybrid compression / decompression unit 502 compresses data by first segmenting the data into regions, and then classifying each region as either a lossless region or a lossy region, so that the appropriate compression algorithm can be applied to each type of region. A lossy compression algorithm, operating according to specified compression parameters, produces an encoded bitstream, which is passed to the compression memory management unit 504 to be arranged into data blocks grouped in such a way as to maintain sustained image quality. Each group of data blocks forms a spectral partition.

[0070] All partitions are classified according to the visual significance of the information they represent. The compression memory management unit 504 also determines whether partitions should be deleted to meet a memory quota. It maintains a minimum level of image quality by limiting the number of partitions that can be deleted. During the hybrid compression process, the hybrid compression / decompression unit 502 also sends metadata relating to the compression data size and compression parameters to be collected and stored by the recompression unit 503.

[0071] After the initial compression pass is complete, the compression memory device 504 then decides whether further action is needed to reduce the amount of compressed data, taking into account the storage requirement (or storage capacity) and the limitations of the partition deletion strategy.

[0072] When the compression memory management unit 504 determines that recompression is needed to reduce the size of encoded data, it initiates the recompression process by first informing the recompression management unit 503 of the amount of memory that needs to be reduced to meet the memory requirement. The recompression management unit 503 then uses the metadata it collected from the initial compression pass to determine how best to adjust the compression parameters to achieve the target memory allocation. The recompression management unit 503 then instructs the hybrid compressor / decompression unit 502 to perform decompression, which reads the encoded data from the compression memory management unit 504 and recompresses the encoded data using the adjusted compression parameters.The output of the RIP output compression device 204 is the compressed image data 505 that meets the storage requirement. Compression process

[0073] Fig. Figure 6 shows a schematic flowchart of an example compression process 600 for implementing the RIP output compression device 204. The disclosed example of the RIP output compression device 204 guarantees that the amount of compressed image data 505 meets the storage requirement by applying a two-stage process as shown in Figure 600. The first stage of the process reads the image data in step 601 and performs initial compression in a subsequent step 602. The second stage of the process reduces the size of the encoded data, if necessary, as determined by a decision step 604, by employing strategic storage reduction in step 603, before the entire compression process concludes in step 605. If a reduction in the size of the encoded data, as determined by step 604, is not required, the process 204 follows a "yes" arrow to step 605.

[0074] Each stage of the compression process 204 is described in more detail below. The hybrid compression / decompression unit

[0075] Fig. Figure 7 shows an example of the hybrid compression / decompression unit 502, which is used for initial image compression in step 602 and as image recompression, if required, as part of the strategic memory reduction in step 603.

[0076] The input image data 501 or the decompressed image data 710 are first processed by a pixel sequence generation unit 701, in which consecutive pixels on the same scanning line of the same color are joined to form a pixel sequence 702. The pixel sequences 702 are then processed by an edge generation unit 703, in which pixel sequences of the same color as a neighbor (by 8-way join as detailed below with reference to Fig. (described in 11) are joined together on successive scan lines to form edge pairs 704. Each edge pair 704 is then evaluated by an edge classification device 705 using an edge evaluation process, which is described in more detail below with reference to Fig. 12. Suitable or valid edge pairs, that is, edge pairs that meet the edge evaluation criteria, are retained in an edge memory 714. The other (invalid) edge pairs are discarded. The edge classifier 705 also sends metadata containing information describing the valid edge pairs and the edge evaluation criteria used in the edge evaluation process, to be collected by the recompression management unit 503 for the purpose of recompression. The valid edge pairs 704 retained in the edge memory 714 are compressed by the lossless compression unit 706. Before being discarded by the edge classifier 705, the invalid edges are sent to a bitmask generation unit 707, where a bitmask is generated on the invalid edges to identify the regions selected for lossy compression.Therefore, lossy regions are those defined by discarded edge pairs. Before the lossy regions are sent to the lossy compression algorithm, the gaps between the lossy regions, identified by the bitmask, are filled with the same colors in a process called padding, performed by a padding device 708, to minimize visual artifacts produced by the lossy compression device 709.

[0077] The resulting compressed data bitstreams 712 and 713 from the lossless compression unit 706 and the lossy compression unit 709, respectively, are passed to the compression memory management unit 504 to be arranged in quality partitions according to their visual significance. The partitions are ranked so that consistent image quality is maintained.

[0078] The compression memory management unit 504, together with the recompression management unit 503, performs a strategic memory reduction in step 603, performing one of the following three actions to reduce the size of compressed data if the memory requirement is not met: a) Deleting quality partitions starting with the least visually significant partition, OR b) Adjusting the parameters of the edge classification unit 705 before recompressing the data using the hybrid compression / decompression unit 502, and then, if necessary, deleting quality partitions, OR c) Changing a compression mode to bypass the lossless compression device 706 and recompressing the data using only the lossy compression device 709, and then, if necessary, deleting quality partitions.

[0079] The strategic memory reduction process 603 and the edge evaluation criteria are described in more detail below, each with reference to Fig. 18 and Fig. 12 described. Pixel sequence generation

[0080] Raw pixel data 501 is fed to the pixel sequence generation device 701 by the RIP 203. For the purposes of this description, pixel data is fed by the RIP 203 on a tile-by-tile basis, with pixels within a tile being fed in tile grid order. In the disclosed arrangement, pixel values ​​are represented by three channels, representing a red component (R), a green component (G), and a blue component (B) (collectively referred to as RGB), with 8-bit accuracy per channel (also known as bit depth). Other color representations, such as a cyan, a magenta, a yellow, and a black channel, can also be used, along with other bit depths.

[0081] Each pixel value within a tile is checked by the pixel sequencer 701 in pixel grid order. The output of the pixel sequencer 701 is the pixel sequences 702. For the purposes of this description, a pixel sequence consists of a sequence of consecutive pixels, all contained within a single tile row, that have substantially identical color values. Preferably, the color values ​​in a sequence are identical, that is, the tolerance between colors is zero. An example of a pixel sequencer is described in more detail below with reference to Fig. 8 described.

[0082] Fig. Figure 8 shows a schematic flowchart of an example of a Process 800 executed by the Pixel Row Generator 701. On a tile-by-tile basis, the Pixel Row Generator 701 generates pixel rows. Pixel row information is stored in a suitable data structure that includes at least one color value for the pixel row and a pixel row length counter to store the length of the pixel row.

[0083] With reference to Fig. In step 801, received input pixels are read one after the other to obtain the next pixel. A subsequent pixel similarity decision step 802 then determines whether the current pixel color value P[X, Y] is identical or substantially identical (as represented by a symbol “==” in the figure) to a previous pixel color value P[X-1, Y]. If the current pixel color value P[X, Y] is identical or substantially identical to the previous pixel color value P[X-1, Y], then process 800 follows a yes arrow from step 802 to step 803, where the end-of-row check step 803 determines whether the tile sample line (tile line) is complete.If the tile sampling line is not yet finished, then process 800 follows a no arrow to step 804, in which the pixel row length counter is incremented in the length increment step 804, and process 800 then returns to step 801 of obtaining a next pixel, from where the next pixel in the tile sampling line is processed.

[0084] If in pixel similarity step 802 it is determined that the current pixel color value P[X, Y] is not identical to the previous pixel color value P[X-1, Y], or if in step 803 it is determined that the tile sampling line is finished, then process 800 is directed to step 805, in which the pixel row is finished and its contrast with neighboring pixels is calculated in the contrast calculation step 805.

[0085] After the contrast has been calculated, process 800 is directed to step 806, which sends the pixel row to the edge generation unit 703 in the send pixel row step 806. This means that the pixel row generation unit 701 only needs to store one pixel row at any given time during the processing of the tile.

[0086] The contrast calculation step 805 determines the contrast of a pixel array using the amount of color difference between the pixel array and its surrounding pixels.

[0087] Fig. Figure 9 shows an exemplary pixel array consisting of a single pixel X (identified by a reference sign 901). Contrast can be determined as the sum of the magnitudes of the differences between pixel array 901 and all its neighboring pixels 902 to 909 for each color channel. Alternatively, contrast can be calculated using a subset of its neighboring pixels. In the disclosed arrangement, the contrast-calculating step 805 computes two contrast values, designated Contrast_L and Contrast_R. Contrast_L is determined as the sum of the magnitudes of the color difference for each color channel between pixel array 901 and pixel 909, that is, the pixel immediately to its left. Contrast_R is determined as the sum of the magnitudes of the color difference for each color channel between pixel array 901 and pixel 905, the pixel immediately to its right. Edge joining criteria

[0088] With reference to Fig. The edge-generating device 703 receives the pixel rows 702 to create regions of connected pixel rows within a tile that have substantially the same color values. Edges mark the boundaries between adjacent regions of such connected pixel rows. Each region requires a pair of edges to fully describe its boundary. An "activating" edge represents the left side of a region, and a "deactivating" edge represents the right side of a region. For the purpose of this description, the activating and deactivating edges are referred to as an "edge pair."

[0089] An edge generation process 1100, as detailed below with reference to Fig. The process described in Figure 11, which is executed by the edge generation device 703, generates the edge pairs 704 that connect rows of pixels with identical color values ​​on successive scan lines, forming areas of identical (or substantially identical) color as described above. New edge pairs, when generated, are considered active until excluded from propagation. The edge pairs 704 are extended when a row of pixels on a current scan line overlaps an active edge pair and meets the joining criteria. For a row of pixels to join an active edge pair, the row of pixels must (a) overlap (using 8-way connectivity) an area currently spanned by an active edge pair, and (b) have a color value identical to that associated with the edge pair. As detailed below with reference to Fig. As described in section 10, it is advantageous to consider active edge pairs on a previous sample line when attempting to determine whether a row of pixels on a current sample line joins any existing active edge pairs. Edge pairs 704 are not allowed to cross other edge pairs, and the planar areas described by edge pairs 704 within a tile are separate. Edge pairs 704 can be excluded from continuation in two ways: (i) a row of pixels on the next tile sample line spans an active edge pair in such a way that the active edge is excluded from continuation, or (ii) the last sample line in the tile is processed and the tile ends. In the case that the edge is prevented from being continued, it is considered "resolved" and marked as inactive.

[0090] Fig. Figure 10 shows examples 1001, 1002, 1003, 1004, 1005, and 1006 of edge pairs 704 where the joining criteria are met. In each example, a row of pixels on the current sample row joins an active edge pair on a previous sample row if the color value of the pixel row on the current sample row is substantially identical to the color value of the active edge pair on the previous sample row. Examples 1001 through 1006 illustrate scenarios where an active edge pair is joined to a pixel row using the 8-way connectedness rule. Edge generation

[0091] Fig. Figure 11 shows a flowchart of an example of process 1100, which is carried out by the edge generation unit 703. The edge generation unit 703 carries out process 1100 until the entire tile has been processed.

[0092] Process 1100 begins in a pixel row read step 1101, in which the next pixel row is read from the pixel row generation device 701. If a subsequent check-if-first-row step 1102 determines that the pixel row occurs on the first sample row of a tile, process 1100 proceeds, according to a yes arrow, to an edge start step 1103, in which a new edge pair is generated. This edge pair is marked as active. After edge start step 1103, process 1100 proceeds to step 1112, and the process ends.

[0093] Alternatively, if the Check-If-First-Row step 1102 determines that the pixel row occurs on a subsequent row in a tile, process 1100 follows a No arrow, and the pixel row is checked in a subsequent Connectivity-Check step 1104 to determine whether the pixel row can join any existing active edge pairs. If step 1104 determines that the pixel row cannot join any of the existing active edge pairs, the Edge Generation Unit 703 proceeds, following a No arrow, to an Edge-Start step 1105, in which a new edge pair is generated and set to active, after which process 1100 proceeds to a Resolve-Check step 1110.Alternatively, if the Connectivity Check step 1104 determines that the pixel array can join an overlapping active edge pair, the Edge Generation Unit 703 proceeds, according to a Yes arrow, to an Edge Extension step 1106, in which the active edge pair is extended by the pixel array. After Edge Extension step 1106, the Edge Generation Unit 703 proceeds to the Resolve Check step 1110.

[0094] In the resolve-check step 1110, it is determined whether the pixel sequence extends past other active edge pairs within the tile, thereby preventing them from being continued. If so, process 1100 follows a "yes" arrow, and the affected edge pairs are marked as "resolved," meaning they are set as inactive in a subsequent edge-resolve step 1111 before process 1100 proceeds to step 1112 and terminates. Alternatively, if the pixel sequence does not resolve any active edge pairs in step 1110, process 1100 proceeds to step 1112 and terminates.

[0095] For the purpose of this description, edge pairs 704 generated by the edge generation device 703 are sorted in the ascending y-direction and then in the ascending x-direction by the starting position of the activating edge. The area bounded by the edge pair 704 has a uniform color. Edge assessment

[0096] The edge classifier 705 evaluates resolved edge pairs 704 generated by the edge generator 703 with respect to visual significance and determines whether an edge pair is valid or not. An edge pair deemed visually significant is considered valid, and an area bounded by a valid edge pair is considered an "area" area. An area is considered area-wide, meaning that the area has a uniform color. Only edge pairs describing an area-wide region are retained and losslessly compressed by the lossless compression unit 706. The areas defined by invalid edge pairs are identified as lossy areas by the bitmask generated by the bitmask generator 707.The invalid edge pairs are discarded, and the masked lossy areas are filled by the padding device 708 and then passed to the lossy compression device 709 for lossy compression.

[0097] In the described SPR arrangements, the edge classifier 705 uses a combination of two criteria to determine whether an edge pair is valid or not. These criteria are as follows: (i) the maximum width of the area defined by the edge pair and the edge pair length (the length of the area defined by the edge pair), and (ii) the contrast significance of the area defined by the edge pair, in order to assess the visual significance of an edge pair.

[0098] The above criteria are collectively referred to as the edge assessment criteria.

[0099] Other edge evaluation criteria can also be used, such as the area defined by the edge pair.

[0100] Fig. Figure 12 shows a schematic flowchart of an example 1200 of an edge evaluation process and a subsequent bit masking process, performed by the edge classification device 705 and the bit mask generation device 707. The process 1200 begins, upon receiving an edge pair generated by the edge generation device 703, with an edge contrast check step 1203, in which it is determined whether the edge pair is significant with respect to contrast. In the disclosed arrangement, an edge pair is considered significant with respect to contrast if more than half of the pixel rows forming the area bounded by the edge pair are judged to have high contrast according to a certain contrast threshold CT_Threshold, using the result calculated in step 805 performed by the pixel row generation device 701.

[0101] Edge pairs deemed significant with respect to contrast by step 1203 cause process 1200 to follow a "yes" arrow, so that the edge pairs are retained in the edge storage device 714 by a subsequent edge pair retention step 1204. The metadata of the edges in the retained edge pairs is sent to the recompression management device 503 for storage by a subsequent step 1205, before process 1200 proceeds to a finish step and the process terminates.

[0102] For retained edge pairs, the edge metadata can be used to determine the number of bytes required to encode the edge pair using a predetermined entropy encoding procedure to be executed by the 706 lossless compression unit. The number of bytes required to encode each edge pair can be accurately predicted before the edge pair is entropy-encoded by the 706 lossless compression unit. The entropy encoding procedure and a size prediction for encoded edge pairs are described in detail below.

[0103] Alternatively, if the area defined by the edge pair is deemed not significant with respect to contrast, process 1200 proceeds from step 1203 to a size significance calculation step 1201, as indicated by a "no" arrow. In this step, an area size significance metric M is calculated for an edge pair evaluation. Process 1200 considers the y-direction length of the edge pair (the length of the area defined by the edge pair), stored in the edge pair data structure as `edge_length`, along with the maximum width of the area defined by the edge pair, stored in the edge pair data structure as `max_width`. These two values ​​are added together to form the area size significance metric M for the edge pair. Typically, for an edge pair to be accepted, the value of the area size significance metric M should be greater than approximately 8 to 10% of the tile's perimeter.For example, in a tile with dimensions of 64 pixels by 64 pixels, an acceptable threshold for the metric M would be 20 (pixels). Accordingly, a subsequent size significance check step 1202 determines whether the range size significance metric M for the edge pair is greater than or equal to a specific edge significance threshold ES_Threshold, which is a range size threshold. If the range size significance metric M for the edge pair is determined to be greater than or equal to ES_Threshold, the edge pair is considered significant with respect to size and is retained in the subsequent edge retention step 1204, after the process follows a yes arrow from step 1202 to step 1204. The edge metadata is then sent by step 1205 to the recompression management facility 503 to be stored before process 1200 terminates. Value candidates for ES_Threshold are obtained from a memory such as an HDD 2110 or a memory 2106.

[0104] Edge pairs that are not significant with respect to contrast in the Edge Contrast Check step 1203 and have an area size metric M smaller than the ES_Threshold as judged by step 1202 are considered visually insignificant and are passed to the Edge Update Mask step 1206 after the process follows a no arrow from step 1202 to step 1206. The Bitmask Generating Unit 707 updates the tile bitmask in the Edge Update Mask step 1206 to reflect the pixel positions of the area enclosed by the visually insignificant edge pair. The edge pair data is then discarded in a subsequent Edge Discard step 1208 before the Edge Evaluation Process 1200 proceeds to the Finish step and terminates. Edge metadata

[0105] As in Fig. As shown in Figure 19, the edge metadata can be used to map a value of the edge significance threshold ES_Threshold to the sum of the coded data sizes for all valid edge pairs at the specified ES_Threshold. Specifically, it is determined that if the size of the area bounded by the edge pair exceeds the edge significance threshold ES_Threshold, a data set is estimated for lossless encoding of edges (edge ​​pairs) that form the area exceeding the edge significance threshold ES_Threshold.

[0106] In this way, the size of the encoded data of an edge pair can be determined using a predetermined entropy encoding method before the edge pair is entropy-encoded by the lossless compression device 706. The entropy encoding method is described in more detail below.

[0107] Mapping between different candidate values ​​of the edge significance threshold ES_Threshold and the encoding data size of all valid edges can be used to accurately predict the memory allocation of all losslessly encoded edge pairs when the ES_Threshold value is adjusted for recompression purposes.

[0108] Fig. Figure 13a shows an example of the edge metadata stored in the recompression management unit 503, using a table 1300 to map each possible value of the edge significance threshold ES_Threshold to the total number of bytes required to encode all valid edges, that is, edges judged to be significant either with respect to contrast in step 1203 or significant with respect to size in step 1202. Table 1300 shows examples of possible ES_Threshold values ​​for a tile with dimensions of 64 pixels by 64 pixels, with table entry 1301 showing the minimum ES_Threshold value above its default value of 20, and table entry 1303 showing the maximum ES_Threshold value.

[0109] Table 1300 does not explicitly list all ES_Threshold candidate values ​​stored in this example. An intermediate ES_Threshold value is also shown in table entry 1302. As the size significance threshold ES_Threshold increases, the total coding size for all valid edges decreases due to a drop in the total number of size-significant edge pairs. It should be noted that, since edge pairs that are significant with respect to contrast are retained in the edge evaluation process 1200 regardless of their size significance metric M, the number of contrast-significant edge pairs does not change when the ES_Threshold value changes.

[0110] Fig. Figure 13b shows a table 1390 of edge metadata stored in the recompression management unit 503, which is an alternative to table 1300. In table 1390, only a subset of all possible values ​​of the edge significance threshold ES_Threshold is selected. Each selected ES_Threshold value is plotted against the total number of bytes required to encode all edges valid according to that threshold. Selecting a subset of all possible edge significance threshold values ​​reduces the memory management data required to retain edge metadata in the recompression management unit 503, as in the example in Figure 13b. Fig. Figure 13b shows that the number of table entries in Table 1390 is reduced compared to the number of table entries in Table 1300. Lossless compression device

[0111] After all edge pairs have been processed by the edge classifier 705, the edge compression unit 706 compresses the visually significant edge data (that is, the information regarding valid edge pairs) stored in the edge memory 714 using a lossless compression method, producing a lossless bitstream 712 which is sent to the compression memory management unit 504 ( Fig. 7) is handed over.

[0112] All data elements of visually significant edge pairs in the disclosed SPR arrays consist of their raw color value, (X, Y) start positions, and enabling and disabling edges. The raw color values ​​corresponding to each edge pair are written to the compression data storage management facility 504 in the same order in which the edge pairs were generated in the tile. There is a one-to-one relationship between the number of raw color values ​​and the number of visually significant edge pairs.

[0113] Since the edge pairs are tile-based, a fixed number of bits can be used to encode the (X, Y) starting positions. Similarly, a fixed number of bits can be used to encode the length of the edge pair in the y-direction, indicating the number of pixel rows that comprise the edge pair. For example, in a tile size of 64 by 64 pixels, 6 bits are needed to encode each coordinate value in the starting positions and length of an edge pair.

[0114] The position of the activating edge in an edge pair is encoded as a series of x-offset values, where each x-offset value represents the change in the x-position of the activating edge from the previous sample line. The deactivating edge is encoded in the same way as activating edges. The offset data can be encoded using an entropy-based encoding method similar to that used by JPEG to encode DC coefficients. In this method, each x-offset value is represented by a symbol pair, referred to as Symbol-1 and Symbol-2. Symbol-1 represents the number of bits in the offset value. Symbol-2 represents the offset value itself. Symbol-1 is encoded with a variable-length code generated from a predetermined suitable Huffman table. Symbol-2 is a variable-length integer whose length in bits is represented by Symbol-1.The Huffman codes and code lengths are determined externally and are known to both the lossless compression unit 706 and the decompression unit 711.

[0115] Since the size of symbol-1 and symbol-2 is known for each x-offset value, the encoding size of an edge pair can be calculated by the edge classifier 705 by summing the sizes of symbol pairs for all x-offsets of the edge pair. The encoding size then forms part of the edge metadata that is sent to the recompression management unit 503 in step 1505.

[0116] If a tile contains only visually significant edge pairs, then the activating edge of one planar region serves as the deactivating edge of an adjacent planar region. Therefore, only the activating edge of each edge pair needs to be compressed by the lossless compression device 706, and the encoding size of an edge pair (size of an encoded edge pair) would only contain the size of x-offsets from the activating edge. Lossy compression and quality partitions

[0117] The lossy areas of a tile, other than the filled areas where edges are to be compressed losslessly, are compressed using a lossy compression algorithm executed by the lossy compression unit 709, after the filled areas, each of which has a uniform color, have been hidden and filled. JPEG is one of the most popular lossy encoding methods for dot-matrix images, employing frequency domain encoding.

[0118] Fig. 14a and Fig. Figures 14b are relevant to this process. In particular, Fig. 14a a schematic diagram of an image divided into coding units for lossy coding, and Fig. Figure 14b is a schematic diagram of an 8 by 8 DCT block.

[0119] During a typical JPEG compression process, a dot-matrix image 1401 is first divided into encoding blocks 1411 of 8 by 8 pixels. A discrete cosine transform (DCT) is applied to each block to generate an 8 by 8 block of frequency domain coefficients 1405, with the DC coefficient 1415 arranged at position (0, 0), and AC coefficients arranged in a zigzag order 1425. The higher the order of the AC coefficients, the higher the frequency of the energies they encode. In the spatial domain, high-frequency energies within the image represent rapidly changing values, and low-frequency energies represent slowly changing gradients.

[0120] After the DCT transformation step, the 64 DCT coefficients are then quantized using a quantization table. Typically, the quantization table will quantize higher-order AC coefficients to a greater degree. This is because the radio-frequency energies they encode can typically be represented with a smaller number of bits. Finally, the quantized DCT coefficients are Huffman-coded.

[0121] Fig. Figure 15 is a functional block diagram of an example lossy compression device 709 that implements a JPEG algorithm as described above. First, a color conversion device 1501 converts input image data into the correct processing color space, if necessary, in this case RGB. Alternatively, input image data can also be converted into a YCbCr color space. Then, a DCT unit 1502 performs a DCT on this RGB data in units of 8 by 8 blocks and outputs DCT coefficients. A quantizer 1503 quantizes the 8 by 8 blocks of DCT coefficients using a modified JPEG quantizer table 1512, thereby outputting quantized coefficients. The quantizing device 1503 also rearranges the 8 x 8 blocks of quantized coefficients into 64 one-dimensional coefficients from low-frequency components to higher-frequency components according to a sampling in a zigzag order as in Fig. 14 shown.

[0122] The quantized coefficients are then partitioned into quality partitions according to a partition table 1514 by a partitioning device 1504, and the partitioning device 1504 feeds the quality partitions to a Huffman coding device 1505, which encodes the input one-dimensional data in the quality partitions using a Huffman table 1513 and a known standard Huffman coding technique. The outputs of the Huffman coding device are the encoded data, where the length (represented by the number of bits, the "code length") of each encoded data item and a number (hereafter referred to as the "partition number") represent a partition to which the correspondingly encoded data items are assigned.

[0123] A partition control device 1511 is part of the compression memory management device 504 and controls the size of the final encoded image by discarding the image's DCT coefficients, starting with the highest-ranking coefficients and ending once the memory limit is met. This is because the majority of the visual information in the image is encoded within the DC and lower AC coefficients.

[0124] Fig. 16a and Fig. Figure 16b describes a method for spectral JPEG coding of images. For each DCT coding block, the quantized coefficients are stored within spectral bands of decreasing visual quality. The increasing order of the DCT coefficients directly relates to their decreasing influence on the visual quality of the image. Thus, the DCT coefficients can be grouped into spectral partitions sorted according to decreasing relevance to the overall image quality. In this arrangement, the image quality can be progressively reduced by deleting the higher spectral partitions and their associated DCT coefficients.

[0125] Fig. Figure 16a shows a partitioning example of the DCT data 1610 for a coding block with partitions 2 and 3 shown as 1615 and 1625.

[0126] Fig. Figure 16b is a table showing an example partitioning scheme. In an SPR arrangement, partition 1 is reserved for lossless encoding data of the edge pairs.

[0127] Based on the "partition number" determined by the compression memory management unit 504, the partition control unit 1511 writes compressed data to a compression memory 1599 according to a partition indicated by the partition number. This compression memory is organized into partitions (see Fig. 20), and each piece of encoded data is written to a partition according to the partition number of that data. Example of edge assessment and area classification

[0128] Fig. Figure 17 shows an example of a tile that is hybrid compressed according to the SPR arrangement. An 8-pixel by 8-pixel tile 1700 is shown. For the purpose of this description, the hybrid compression / decompression unit 502 sets an ES_Threshold value of 3. The pixel array generation unit 701 generates six pixel arrays 1701 to 1706 for a first sample line. The edge generation process 1100 receives each pixel array and accordingly generates one active edge pair for each pixel array. No edge pairs are evaluated during the processing of the first sample line because all edges are still active. For simplicity, only the edge pairs 1715 and 1716 for pixel arrays 1703 and 1706, respectively, are shown. Fig. 17 shown.

[0129] On the second sampling line, pixel row 1707 extends the active edge pair 1715, which was started by pixel row 1703. The x-offset value for both the activating and deactivating edges of the active edge pair 1715 on the second sampling line is 0. Pixel rows 1718, 1719, 1720, 1721, and 1708 on the second sampling line start new active edges, as they cannot extend edge pairs from the first sampling line. After pixel row 1708 is processed by edge generation process 1100, the edge pairs formed by pixel rows 1701, 1702, 1704, and 1705 are all resolved by edge generation process 1100, as they can no longer be extended. Each of these resolved edges has low contrast with respect to the specified contrast threshold CT_Threshold and a magnitude significance metric M of 2.Thus, in edge evaluation process 1100, the four edge pairs formed by pixel rows 1701, 1702, 1704, and 1705 are discarded. Pixel row 1709 extends edge pair 1716, which was started by pixel row 1706, to include pixel row 1709. The x-offset value for activating and deactivating edges of edge pair 1716 on the second sampling row is +1 and 0, respectively.

[0130] Edge generation process 1100 continues for the remainder of the tile. Edge pair 1716 is extended with pixel rows 1712, 1713, and 1714. Edge pair 1715 is extended with pixel row 1711. The remaining pixels mapped by 1717 are considered part of the image area, and their corresponding edge pairs are discarded.

[0131] After all edge pairs have been classified, the tile is compressed using (a) the lossless compression device 706 to compress edge pairs 1715 and 1716, and (b) the lossy compression device 709 to compress the rest of the tile, with the planar areas hidden and filled with colors from neighboring pixels, e.g. pixels 1704, 1705 and 1708. Strategic storage reduction

[0132] Referring to Fig. 5 and Fig. 6. After an initial compression pass in step 602 is performed by the hybrid compression / decompression unit 502, control is transferred to the compression memory management unit 504, where step 604 is performed to determine whether further actions are needed to reduce the size of the compressed data to meet the target memory size. If so, step 603, a strategic memory reduction step, is performed, which is described in more detail below with reference to the schematic flowchart in Fig. 18 is explained.

[0133] Fig. 18 is a schematic flowchart that illustrates the process described in Fig. Figure 6 illustrates a step in a strategic memory reduction process. The strategic memory reduction process 1800 is triggered by a decision made by the compression memory management unit 504 to reduce the amount of compressed data. In step 1801, the compression memory management unit 504 checks the data sizes of the quality partitions of the lossless compressed image data and determines the minimum memory allocation if the number of retained partitions exceeds a predetermined minimum quality level. The minimum quality level must be maintained to avoid noticeable visual artifacts in the areas of lossily compressed image data. This step precedes an edge evaluation process in Fig. Step 12 (for example, edge contrast determination step 1203 or area size determination step 1201) is performed. If the given memory requirement can be met by retaining at least the number of quality partitions needed to maintain the minimum quality level, a subsequent step 1802 is executed by the compression memory management unit 504 after the process follows a yes arrow from step 1801 to step 1803. Step 1802 removes quality partitions one by one, starting with the partition containing the least visually significant data, until the memory requirement is met.

[0134] Simply removing the least visually significant quality partitions alone might not sufficiently reduce memory usage; that is, a minimum quality level might not be achievable within the constraints of the target memory limit. This scenario can occur particularly when only a small portion of the data is lossily compressed in the initial hybrid compression process, and the majority of the data is losslessly compressed, thus precluding the use of incremental quality partition deletion. To maintain the optimal balance between memory usage and image quality, a strategy of shifting data from lossless to lossy areas during the recompression process is employed.

[0135] If the compression memory management unit 504 determines in step 1801 that the target memory cannot be reached while maintaining the minimum quality level in step 1801, it transfers control to the recompression management unit 503 and instructs it to recompress the image. The recompression management unit 503 then determines the best recompression modes in a subsequent step 1803, after the process follows a "no" arrow from step 1801 to step 1803, based on the edge metadata it collected from the edge classifier unit 705 during the initial hybrid compression pass.The recompression management unit 503 selects between (a) recompression in a hybrid (i.e., mixed) mode with a more aggressive edge significance threshold ES_Threshold, allowing more image data to be classified as lossy areas, and (b) recompression in a fully lossy mode, bypassing the edge generation and classification process and using only the lossy compression unit 709. In either case, recompression will reduce the size of compressed data in the lossless area. Either all data (in the case of fully lossy compression) or a sufficient portion of the image data will be lossily compressed so that the storage requirement can be met by progressively deleting quality partitions.

[0136] Since data recompression is computationally intensive, it is advantageous to determine the appropriate recompression mode that guarantees the resulting compressed data meets the memory requirement in a single pass with optimal image quality. To do this, the recompression management unit 503 must determine whether a suitable value for the edge significance threshold ES_Threshold can be found in step 1803 to ensure sufficient reduction of the memory footprint of the losslessly compressed data. If such a threshold exists, the recompression management unit 503 adjusts the edge significance threshold in a subsequent step 1804, after process 1800 follows a yes arrow from step 1803 to step 1804, and then initiates the hybrid recompression in a subsequent step 1805.

[0137] If not, a fully lossy recompression process 1806 is initiated after process 1800 follows a "no" arrow from step 1803 to step 1806. In the disclosed arrangement, the lossy compression algorithm is the same lossy compression algorithm used as part of the hybrid compression.

[0138] The process of selecting a suitable value from a large number of candidate values ​​for the edge significance threshold ES_Threshold is described in more detail below with reference to Fig. 19 described.

[0139] Fig. Figure 19 shows an example histogram representing the metadata used to adjust the edge evaluation criteria. Histogram 1900 is an alternative view of the example metadata table 1390 collected by the recompression management facility 503, where the x-axis shows the range of valid ES_Threshold (range size threshold) values ​​to show the candidate values ​​for this particular example, and the y-axis shows the total number of bytes required to losslessly encode all valid edges that meet each ES_Threshold. An example of a memory requirement is marked by a dashed horizontal line 1905 to indicate the target memory size. The maximum losslessly compressed data size 1906 is also referred to as the target recompression size.The maximum lossless compressed data size 1906 as a storage requirement is determined by subtracting the amount of storage 1903 for storing lossily compressed image data with the predetermined minimum quality partition level from the predetermined target storage size (or storage capacity) 1905.

[0140] In this particular example, all estimated data sizes (the total number of bytes required to encode all valid edges reaching ES_Threshold) are compared to the maximum lossless compressed data size of 1906 as a memory requirement. It can be seen that the total number of bytes required to encode all valid edges reaching an ES_Threshold value of 26 (i.e., reference 1901) is insufficient to meet the memory requirement at the minimum quality level. Therefore, the next reference value, 32, is selected as the new value for the edge significance threshold, ES_Threshold.The total number of bytes required to encode all valid edges according to the selected edge significance threshold (ES_Threshold) is closest to the maximum lossless compressed data size (1906) and within the maximum size of 1906 lossless compressed data as a storage requirement. Since the total number of bytes required to encode all valid edges can maintain the highest image quality and also stays within the maximum size of 1906 lossless compressed data, the new ES_Threshold is passed to the edge classifier 705 in step 1804 before the recompression process begins.

[0141] If there is no suitable ES_Threshold value to select, that is, in the case that the maximum allowed lossless compressed data size 1906 is smaller than the lossless encoding size at maximum ES_Threshold value, the recompression management unit 505 will select step 1806 of completely lossy recompression in step 1803, resulting in an encoding size of zero for lossless data.

[0142] Returning to Fig. In either step 1805 or step 1806, a recompression process is performed in which the compressed data is first decompressed tile by tile. After each tile of data has been decompressed by the decompression unit 711, a compression process is performed on the tile data using the compression mode selected by the recompression management unit 503 in step 1803. If a hybrid recompression mode is selected, the compression process is similar to the initial hybrid compression 602, except that a new value for the edge significance threshold ES_Threshold is used to classify edges. Specifically, edges that form the area that meets the selected value of the area size threshold are compressed losslessly. Image data in an area other than the area to be compressed losslessly is also compressed lossily.If full lossy recompression is selected, then the compression process is simplified; that is, instead of sending the tile data to the pixel array generation unit 701, decompressed tile data is sent directly to the lossy compression unit 709 for full lossy compression.

[0143] After the recompression process is complete, the compression memory management unit 504 compares the compressed data size with the memory requirement in a subsequent step 1807. If the compressed data size is less than or equal to the target memory limit, the strategic memory reduction process ends after process 1800 follows a "yes" arrow to a final step. However, if the compressed data size is larger than the required target memory limit, the compression memory management unit receives a notification of this and, in a step 1802, after process 1800 follows a "no" arrow from step 1807 to step 1802, releases memory by deleting quality partitions, starting with the lowest-quality partition, until the target memory limit is reached.In this way, the image quality of the final output is reduced in a controlled manner, only to the extent necessary to meet the target memory limit. Once the target memory limit requirement is met, the strategic memory reduction process ends. Partition arrangement

[0144] Fig. 20a and Fig. Figure 20b shows an example of quality partitions and how recompression is used to move information between partitions so that increasing quality reduction is possible when further memory usage reduction is needed.

[0145] Fig. Figure 20a shows a schematic diagram of an example of a partition arrangement after an initial compression process 602. Fig. Figure 20b shows a schematic diagram of an example of a partition arrangement during the strategic memory reduction process 603, in which hybrid recompression has taken place.

[0146] According to the SPR arrangement, the compression memory management device 504 organizes the memory containing the compressed image into a plurality of header records and memory chunks. Each memory chunk corresponds to a specific quality partition header record. The partition header records store partition information such as the partition number, the length (that is, size) of the partition, the method for accessing the allocated memory chunk, or other additional information as needed. For the purposes of this disclosure, a “memory chunk” means one or more blocks of not necessarily contiguous memory that are joined together to form a logically contiguous segment of memory for the purpose of storing compressed image data.In this example, segment 2001 corresponds to header record 2091 of partition 1, while segment 2003 corresponds to header record 2093 of partition 3.

[0147] In this example, the compression memory management unit 504 organizes the quality partitions during the initial compression process 602 such that all lossless encoded edges produced by the lossless compression unit 706 are stored in memory segment B1, which belongs to partition 1 (i.e., 2001), while the encoded DCT data produced by the lossy compression unit 709 is stored in each partition within its corresponding memory segment from partition 2 (i.e., 2002) to partition 11 (i.e., 2011). An exemplary mapping of the partition number to its corresponding DCT data is shown in Fig. 16b shown.

[0148] Fig.Figure 20b shows a partition layout resulting from hybrid recompression step 1805. During hybrid recompression by step 1805, the compression memory management unit 504 releases all memory segments from 2001 to 2011, thus decompressing the image. Memory is reallocated to store the re-encoded data as recompression step 1805 continues. During this example of hybrid recompression, the compression memory management unit 504 allocates memory segment B1' (i.e., 2051) to partition 1 to store re-encoded edge data from the lossless compression unit 706. However, the size of memory segment B1' required to store the uncompressed edge data is much smaller than the original size of memory segment B1 (i.e. 2001) that was used to store encoded edge data from the initial hybrid compression 602.This is because the edge significance threshold ES_Threshold was adjusted in step 1804 prior to hybrid recompression, resulting in a reduction in the number of valid edges, referred to as adjusted valid edges or adjusted valid edge pairs. However, the consequence of reducing the number of lossless encoded edges is a potential increase in the size of the partitions storing encoded DCT data from the lossy compression unit 709, due to the correspondingly increased number of invalid edges, referred to as adjusted invalid edges or adjusted invalid edge pairs. As shown in this example, the memory segment B2' (i.e., 2052), allocated to partition 2 used to store the encoded DCT data from the lossy regions, is now slightly larger than the original size of memory segment B2 (i.e., 2052).2002) prior to hybrid recompression. Similarly, all lossy partitions from partition 3 to partition 11 might potentially gain slightly more memory usage due to the increase in the number of lossy areas. The reduction in memory usage through lossless encoding can outweigh the increase in memory usage through lossy encoding. Furthermore, as shown in this example, shifting more edges to lossy compression can free up more memory by deleting partitions with less visually significant DCT data, and less memory is needed to store the losslessly encoded edges and the more significant DCT data in the lower partitions, thus meeting the target memory limit requirement without significant degradation of image quality.

[0149] Under extremely limited storage conditions, the strategic storage reduction process 603 will select to recompress an image in step 1806 using fully lossy mode. In this case, partition 1 has a size of zero, and all image data is encoded in partitions 2 through 11, which can then be gradually deleted in step 1802 until the storage requirement is met. Industrial applicability

[0150] The described arrangements are applicable in the computer and data processing industries.

[0151] The foregoing describes only some embodiments of the present invention, and modifications and / or alterations may be made to them without departing from the scope and spirit of the invention, the embodiments being explanatory and not limiting.

[0152] A method for compressing an image to be stored in memory to meet a storage requirement is disclosed. The size of a region with a uniform color (1201) in the image is determined. The determined region size is compared with the candidate values ​​of the region size threshold (1202). A value is selected from the candidate values ​​as the region size threshold based on a comparison between the estimated data amounts required to meet the region size threshold and the storage requirement. The edges defining the region that meets the selected value of the region size threshold are compressed without loss.

Claims

[1] A method of compressing an image to be stored in a memory to meet a storage requirement, the method comprising the steps of: Determining a size of an area in the image, the area having a uniform color (1201), Obtaining candidate values of a range size threshold (1300, 1390, ES_Threshold), characterized by Comparing the determined size of the area with at least one of the value candidates of the area size threshold (1202), Estimating data sets (1901, 1902, 1904) for encoding edges that define the range that satisfies at least one of the value candidates of the range size threshold, Selecting one value from the candidate values as the range size threshold based on the comparison between the estimated data amounts and the storage requirement (1903), lossless compression of edges defining the region that meets the selected value of the region size threshold (1805), and lossy compression of image data in an area excluding the area to be compressed losslessly (1805). [2] The method of claim 1, wherein the region is defined by the edge pair and the size of the region is determined based on the sum of a length of the edge pair and a maximum width of the edge pair. [3] The method of claim 1, wherein the storage requirement is defined by subtracting the amount of memory for storing lossy compressed image data with a predetermined quality level from a predetermined target storage size. [4] The method of claim 1, wherein the selected one of the candidate values corresponds to an estimated data set that is closest to the storage requirement and satisfies the storage requirement. [5] The method of claim 1, wherein the area whose size is determined does not satisfy a predetermined contrast threshold. [6] The method of claim 1, further comprising the steps Determining whether the number of partitions retained in lossy compressed image data stored in the memory is above a predetermined minimum quality level before determining the size of a region, and Deleting a partition with the lowest quality to meet the storage quota. [7] A method of compressing an image to be stored in a memory to satisfy a storage condition, the method comprising the steps of: Obtaining candidate values of a size of an area to be compressed in a first way, the area having a uniform color in the image, Estimating at least one set of compressed data to be obtained by compressing, in the first way, regions in the image that satisfy one of the value candidates, Selecting a value of a range size from the candidate values based on the estimate and the storage requirement, and Compressing, in the first way, a region of the image that satisfies the selected value, and compressing, in a second way, a region of the image that does not satisfy the selected value, wherein the compression in the first way is less lossy than the compression in the second way. [8] The method of claim 7, wherein the estimating step comprises Determining regions where each edge size exceeds one of the candidate values, and Estimating a data amount of compressed data of the specified regions when the specified regions are compressed in the first way. [9] An apparatus for compressing an image to be stored in a memory to achieve a storage capacity, the apparatus comprising: a processor (2105), and a memory (2109) storing a computer-executable program configured to instruct the processor to perform a method comprising the steps of: Determining a size of an area in the image, the area having a uniform color, Obtaining candidate values of a range size threshold, characterized by Comparing the determined size of the area with at least one of the value candidates of the area size threshold, Estimating data sets for encoding edges that define the range that satisfies at least one of the value candidates of the range size threshold, Selecting one value from the candidate values as the range size threshold based on the comparison between the estimated data amounts and the storage requirement, lossless compression of edges defining the region that meets the selected value of the region size threshold, and lossy compression of image data in an area excluding the area to be compressed losslessly. [10] An apparatus for compressing an image to be stored in a memory to achieve a storage capacity, the apparatus comprising: a processor (2105), and a memory (2109) storing a computer-executable program configured to instruct the processor to perform a method of compressing an image to be stored in a memory to satisfy a storage condition, the method comprising the steps of: Obtaining candidate values of a size of an area to be compressed in a first way, the area having a uniform color in the image, Estimating at least one set of compressed data to be obtained by compressing, in the first way, regions in the image that satisfy one of the value candidates, Selecting a value of a range size from the candidate values based on the estimate and the storage requirement, and Compressing, in the first way, a region of the image that satisfies the selected value, and compressing, in a second way, a region of the image that does not satisfy the selected value, wherein the compression in the first way is less lossy than the compression in the second way. [11] The apparatus of claim 10, wherein the program is further configured to instruct the processor to perform the estimation step by a method comprising the steps of Determining regions where each edge size exceeds one of the candidate values, and Estimating a data amount of compressed data of the specified regions when the specified regions are compressed in the first way. [12] A computer-readable non-volatile storage medium storing a computer-executable program adapted to cause a processor to perform a method of compressing an image to be stored in a memory to satisfy a storage condition, the method comprising the steps of: Determining a size of an area in the image, the area having a uniform color (1201), Obtaining candidate values of a range size threshold (1300, 1390, ES_Threshold), characterized by Comparing the determined size of the area with at least one of the value candidates of the area size threshold (1202), Estimating data sets (1901, 1902, 1904) for encoding edges that define the range that satisfies at least one of the value candidates of the range size threshold, Selecting one value from the candidate values as the range size threshold based on the comparison between the estimated data amounts and the storage requirement (1903), lossless compression of edges defining the region that meets the selected value of the region size threshold (1805), and lossy compression of image data in an area excluding the area to be compressed losslessly (1805). [13] The computer-readable non-transitory storage medium of claim 12, wherein the region is defined by the edge pair and the size of the region is determined based on the sum of a length of the edge pair and a maximum width of the edge pair. [14] A computer-readable non-volatile storage medium according to claim 12, wherein the storage requirement is defined by subtracting the amount of storage for storing lossily compressed image data having a predetermined quality level from a predetermined storage target size. [15] The computer-readable non-volatile storage medium of claim 12, wherein the selected value of the candidate values corresponds to an estimated amount of data that is closest to the storage requirement and satisfies the storage requirement. [16] The computer-readable non-volatile storage medium of claim 12, wherein the area whose size is determined does not satisfy a predetermined contrast threshold. [17] The computer-readable non-transitory storage medium of claim 12, wherein the program is configured to instruct the processor to further perform the following steps Determining whether the number of partitions retained in lossy compressed image data stored in the memory is above a predetermined minimum quality level before determining the size of a region, and Deleting a partition with the lowest quality to meet the storage quota. [18] A computer-readable non-volatile storage medium storing a computer-executable program adapted to instruct a processor to perform a method of compressing an image to be stored in a memory to satisfy a storage condition, the method comprising the steps of: Obtaining candidate values of a size of an area to be compressed in a first way, the area having a uniform color in the image, Estimating at least one set of compressed data to be obtained by compressing, in the first way, regions in the image that satisfy one of the value candidates, Selecting a value of a range size from the candidate values based on the estimate and the storage requirement, and Compressing, in the first way, a region of the image that satisfies the selected value, and compressing, in a second way, a region of the image that does not satisfy the selected value, wherein the compression in the first way is less lossy than the compression in the second way. [19] The computer-readable non-volatile storage medium of claim 18, wherein the program is further configured to instruct the processor to perform the estimation step by a method comprising the following steps Determining regions where each edge size exceeds one of the candidate values, and Estimating a data amount of compressed data of the specified regions when the specified regions are compressed in the first way.

Citation Information

Patent Citations

  • Mate face brazing for turbine components

    US20120201684A1

  • Single-pass guaranteed-fit data compression using rate feedback

    US20030081845A1

  • Image processing apparatus and method thereof

    US20080144953A1

  • Compound document page data compression

    US6373583B1