Decoding methods and apparatus where the number of characters in the first and remaining strings have different ranges of values.

By optimizing the value range and encoding/decoding method of the point vector index representation string in the whole compression unit, the problem of low coding efficiency in point prediction technology is solved, achieving more efficient data compression and throughput, and applicable to one-dimensional, two-dimensional and higher-dimensional data encoding and decoding.

CN123123569APending Publication Date: 2026-11-17郭萍
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611542424.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2021-11-01
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

In existing point prediction techniques, the value range of the point vector index representation string is set unreasonably, resulting in low encoding efficiency. This is especially true for high-pixel resolution image and video encoding and decoding, where an excessively large reference range increases implementation complexity and cost, leading to a severe throughput bottleneck.

Method used

In the entire compression unit, the value range of the number of the first string is set to be greater than or equal to 0, and the value range of the number of the subsequent strings is set to be greater than or equal to 1. Specific encoding and decoding methods, such as truncating binary code, truncating unary code, truncating zero-order exponent Golomb code, etc., are used to encode and decode the number of strings to optimize encoding efficiency.

Benefits of technology

By optimizing the range of string number values ​​and the encoding/decoding method, the efficiency of point prediction coding was improved, the implementation complexity and cost were reduced, and the throughput and quality of data compression were enhanced.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123123569A_ABST
    Figure CN123123569A_ABST
Patent Text Reader

Abstract

This invention provides a method and apparatus for encoding and decoding in which the number of the first string and the number of the remaining strings in an integer compression unit have different value ranges. When encoding and decoding a group of consecutive dot vector index representation strings, a variable called the string number and its syntax elements, representing the number of dot vector index representation strings in the group, are first encoded and decoded. The value range of the number of the first string in the integer compression unit is greater than or equal to 0, and the value range of the number of the subsequent strings is greater than or equal to 1. This reduces the value range of the number of the subsequent strings, effectively improving the efficiency of dot prediction coding.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] This application is a divisional application of the following original application: The original application was filed on 2021-11-01. Application number of the original application: 2021112834822 Original invention application title: Encoding and decoding method and apparatus where the number of the first and subsequent strings has different ranges of values. Technical Field

[0002] This invention relates to an encoding and decoding system for lossy or lossless data compression, and particularly to an encoding method and apparatus for data compression using point prediction, and a decoding method and apparatus. Background Technology

[0003] As human society enters the era of artificial intelligence, big data, virtual reality, augmented reality, mixed reality, cloud computing, mobile computing, cloud-mobile computing, ultra-high definition (4K) and ultra-high definition (8K) video image resolution, and 4G / 5G communication, ultra-high compression ratio and extremely high quality data compression for various types of data, including big data, image data, video data, and various new forms of data, has become an indispensable technology.

[0004] A dataset is a collection of data elements (e.g., bytes, bits, pixels, pixel components, spatial sampling points, transform domain coefficients).

[0005] When encoding or decoding a dataset, the data elements are usually sorted according to a predetermined rule, i.e., the order is specified, and then the encoding and decoding are performed according to the order.

[0006] When encoding (and corresponding decoding) a dataset arranged in a certain spatial (one-dimensional, two-dimensional, or multi-dimensional) shape (e.g., a one-dimensional data queue, a two-dimensional data file, a frame image, a video sequence, a transform domain, a transform block, multiple transform blocks, a three-dimensional scene, or a sequence of continuously changing three-dimensional scenes), especially two-dimensional or higher-dimensional datasets, the dataset is generally divided into several compressed subsets with a predetermined shape and / or size (i.e., the number of elements). The compressed subsets are then encoded or decoded sequentially in a predetermined order.

[0007] When encoding or decoding a compressed subset, this compressed subset is usually divided into several maximum compressed units with a predetermined shape and / or size (i.e., number of elements). Encoding or decoding is performed sequentially, one maximum compressed unit after another, in a predetermined order. Maximum compressed units arranged in a row from left to right constitute a maximum compressed unit row.

[0008] Within a maximum compression unit, this maximum compression unit is further divided into several sub-units with a predetermined shape and / or size (i.e., number of elements), called whole compression units. Encoding or decoding is performed one by one in a predetermined order, using whole compression units as the unit.

[0009] In short, the above process ultimately divides the dataset into several subsets with a predetermined shape and / or size (i.e., number of elements), called whole compression units. Encoding or decoding is performed one whole compression unit at a time in a predetermined order.

[0010] At any given moment, the compressed subset being encoded or decoded is called the current compressed subset. The largest compressed unit being encoded or decoded is called the current largest compressed unit. The row of the largest compressed unit being encoded or decoded is called the current largest compressed unit row. The integer compressed unit being encoded or decoded is called the current integer compressed unit. The data element being encoded or decoded (sometimes simply called an element) is called the currently encoded data element or the currently decoded data element, collectively referred to as the current data element, or simply the current element. An element consists of N components (usually 1 ≤ N ≤ 5), therefore the dataset, compressed subset, largest compressed unit, and integer compressed unit also consist of N components. The components of an element are also called component elements.

[0011] For example, a compressed subset is a frame of image whose elements, i.e., pixels, are arranged in a rectangular shape with a size (resolution) of 3840 (width) x 2160 (height), consisting of three components: G (green), B (blue), and R (red) components, or Y (luminance), U (Cb) and V (Cr) chrominance components. A frame of image is divided into a maximum compression unit of 128x128 pixels. Each maximum compression unit is further divided into smaller compression units of varying sizes, ranging from 4x4 to 64x64 squares or rectangles.

[0012] When a dataset is divided into compressed subsets, maximum compressed units, and integer compressed units, a predetermined rule for sorting the elements is to first sort the compressed subsets, then sort the maximum compressed units within each compressed subset, then sort the integer compressed units within each maximum compressed unit, and finally sort the elements within each integer compressed unit.

[0013] In other words, when the dataset is finally divided into whole compressed units, a predetermined rule for sorting is to first sort the whole compressed units, and then sort the elements within each whole compressed unit.

[0014] The relationship between the sampling rates of components in a multi-component dataset and an integer compression unit, which are the objects of encoding, is usually represented by the sampling format. Data in which all N components have the same sampling rate and size (i.e., the number of component samples) is called full-sampled data. Data in which N components have different sampling rates and sizes, where N1 components are called principal components, and the sampling rate and size of the remaining N-N1 components are called secondary components, and the sampling rate and size of the secondary components are integer multiples of the secondary components, is called downsampled data. These integer multiples are usually 2x, 4x, 8x, 2x2, 4x2, etc. In full-sampled data, all components are considered principal components, and there are no secondary components. In downsampled data, at least one component is a principal component and at least one component is a secondary component. For example, for an array of two-dimensional data elements including computer-generated images containing graphics and text, a sampling format called 4:4:4 (or simply 444) is often used, where all three components of the dataset have the same sampling rate and size (i.e., the number of component samples). For arrays of another type of two-dimensional data elements, including natural images and videos captured by cameras, a sampling format called 4:2:0 (or simply 420) is typically used. This means that for a dataset (such as an image or video) with a rectangular shape and three components, the sampling rate and size of two components called secondary components (D and E components) are one-quarter of the sampling rate and size of another component called the primary component (F component). In other words, there is a 4:1 downsampling relationship between the primary and secondary components. In this case, one D component D[i][j] and one E component E[i][j] correspond to four (2×2) F components F[2i][2j], F[2i+1][2j], F[2i][2j+1], and F[2i+1][2j+1]. If the resolution of the F component is 2M×2N (2M horizontal component elements and 2N vertical component elements), that is, the F component of the dataset is F = {F[m][n]: m = 0~2M-1, n = 0~2N-1}, then the resolutions of the D component and the E component are both M×N (M horizontal component elements and N vertical component elements), that is, the D component and the E component of the dataset are D = {D[m][n]: m = 0~M-1, n = 0~N-1} and E = {E[m][n]: m = 0~M-1, n = 0~N-1}, respectively. In situations where high quality is also required for the secondary components, a sampling format called 4:2:2 (abbreviated as 422) is often used. This means that for a dataset with a rectangular shape and three components (such as an image or video), the sampling rate and size of the two secondary components (D component and E component) are each half that of the other principal component (F component), which means that there is a 2:1 downsampling relationship between the principal component and the secondary component.In this scenario, in one direction (e.g., horizontally) of a dataset (such as an image or video), a D component D[i][j] and an E component E[i][j] correspond to two (2×1) F components F[2i][j] and F[2i+1][j]. If the resolution of the F components is 2M×N, i.e., the F components of the dataset are F = {F[m][n]: m = 0~2M-1, n = 0~N-1}, then the resolutions of the D and E components are both M×N, i.e., the D and E components of the dataset are D = {D[m][n]: m = 0~M-1, n = 0~N-1} and E = {E[m][n]: m = 0~M-1, n = 0~N-1}, respectively. In images and videos using YUV, YCbCr, or YCgCo color formats, the aforementioned F, D, and E components are typically Y, U, V components, Y, Cb, Cr components, or Y, Cg, Co components, respectively. In images and videos using the RGB color format, the F, D, and E components mentioned above are typically the G, B, and R components, or the G, R, and B components, respectively. When the data is an image or video, the sampling format is often referred to as the chroma format. A chroma format where all components have the same sampling rate is called a full chroma format. A chroma format where some components have a downsampling relationship with other components is called a downsampled chroma format.

[0015] In downsampling formats, the location of a minor component (usually called the minor component location) and its elements correspond to the locations of multiple major components (usually called major component locations, even simply referred to as locations will not cause confusion) and their elements. This one-to-many correspondence is uncertain. To eliminate this uncertainty, a positive major component location and its positive element are usually pre-designated among the multiple locations (such as 2x2 locations in the 420 format or 2x1 locations in the 422 format) and their elements corresponding to a minor component location and its elements (e.g., designating the top-left location and its element among the 2x2 locations and their elements, or the leftmost location and its element among the 2x1 locations and their elements as the positive major component location and its positive element). This serves as a unique, regular major component location and its element that corresponds one-to-one with the minor component location and its elements. There is a one-to-one correspondence between the positive positions of the principal components and the positions of the secondary components. Therefore, the positive positions of the principal components and the positions of the secondary components are collectively referred to as positive positions, indicating that the position is both a positive position of the principal component and a position of the secondary component. There is also a one-to-one correspondence between positive positions and secondary component positions. All positions other than the positive position among the multiple positions corresponding to a secondary component position are called non-positive positions. Similarly, there is a one-to-one correspondence between the positive elements of the principal components and the elements of the secondary components. Therefore, the positive elements of the principal components and the elements of the secondary components are collectively referred to as positive elements, indicating that the element is both a positive element of the principal component and a element of the secondary component. There is also a one-to-one correspondence between positive elements and secondary component elements. All elements other than the positive element among the multiple elements corresponding to a secondary component element are called non-positive elements. On the other hand, in the full sampling format, all positions are considered as positive positions and positive positions of the principal components, and all elements are considered as positive elements and positive elements of the principal components.

[0016] In the case where the data is an array or sequence of two-dimensional data elements in a 420 sampling format, there is one principal component F and two secondary components D and E. The sampling rate and size of the secondary components D and E are one-quarter of those of the primary component F, meaning that there is a 4:1 downsampling relationship between the primary and secondary components, or 2x2:1. A D component element D[i][j] and an E component element E[i][j] correspond to 2×2 F component elements F[2i][2j], F[2i+1][2j], F[2i][2j+1], F[2i+1][2j+1] in a top-bottom-left-right arrangement; The resolution of the F component elements is 2M×2N, that is, the array composed of the F component elements is F = {F[m][n]: m = 0~2M-1, n = 0~2N-1}. The resolution of the D component elements is M×N, that is, the array composed of D component elements is D = {D[m][n]: m = 0~M-1, n = 0~N-1}. The resolution of the E component elements is also M×N, that is, the E component elements form an array E = {E[m][n]: m = 0~M-1, n = 0~N-1}.

[0017] The positive element of the principal component at the pre-specified positive position is F[2i][2j], which is called the positive position of the principal component of the upper left corner type and its positive element; or, The positive element of the principal component at the pre-specified positive position is F[2i+1][2j], which is called the positive position of the principal component of the upper right corner type and its positive element; or, The positive element of the principal component at the pre-specified positive position is F[2i][2j+1], which is called the positive position and positive element of the principal component of the lower left corner type; or, The positive element of the principal component at the pre-specified positive position is F[2i+1][2j+1], which is called the positive position of the principal component of the lower right corner type and its positive element.

[0018] When a dataset is divided into compressed units, a pre-defined sorting rule is to first sort the compressed units, and then sort the elements within each compressed unit.

[0019] One effective method for data compression is string prediction, also known as string matching. String prediction divides the elements of a current compressed unit into strings of variable length. For a current string being encoded or decoded (referred to as the current string), a reference string (also called the reference string, prediction string, or matching string) with the same or similar values ​​is obtained from a set or subset of elements that have completed a predetermined degree of encoding or decoding, called the reference set. For a reference string of the current string, only a few parameters are needed to record the position and / or shape and / or size and / or dimensions of the reference string within the reference set, without needing to record the value of each element in the current string individually. This allows for a complete representation and reconstruction of all elements and their values ​​of the current string whenever needed, thus achieving data compression. The elements of the reconstructed current string are called reconstructed elements, and their values ​​are called the reconstructed values ​​of the current string and its elements. Data compression where the reconstructed value of an element is equal to its original value is called lossless compression. Data compression where the reconstructed value of an element is not equal to its original value is called lossy compression.

[0020] For example, given a string ordered sequentially according to a certain scanning method, if a corresponding reference string can be found in a reference set, then only two parameters are needed: the positional relationship between the first element of the current string and the first element of the reference string, and the string length, to record the position and size of the reference string in the reference set. This eliminates the need to record the value of each element in the current string individually. This allows for a complete representation and reconstruction of all elements and their values ​​whenever needed, obtaining the reconstructed values ​​of the current string and its elements. The number of bits consumed in recording these two parameters is often far less than the number of bits consumed in recording the value of each element in the current string individually, thus achieving data compression.

[0021] In string prediction, the reconstruction operation is typically performed by reading a reference element and copying it to the current element's position. Therefore, string prediction is also called string copying. The throughput of producing the reconstructed element largely depends on the throughput of reading the reference element.

[0022] In string prediction, unpredictable elements (also known as unmatched, mismatched, or non-matched elements) may appear where no reference element can be found in the reference set. The components, principal components, and secondary components of an unpredictable element are called unpredictable components, unpredictable principal components, and unpredictable secondary components, respectively. For unpredictable elements, only their exact or approximate values ​​can be recorded, compressed according to a predetermined method, and reconstructed whenever needed to obtain the reconstructed value of the unpredictable element.

[0023] Commonly used scanning methods in string prediction include: Horizontal raster scanning: Elements within a single compressed cell are arranged one by one along the horizontal direction. After one row is completed, the next row is arranged. All scanning directions within a row are arranged from left to right or from right to left.

[0024] or Horizontal back-and-forth scanning, also known as forward and backward scanning or bow scanning, involves arranging elements one by one in a horizontal direction within a single compressed cell. After arranging one row, the next row is arranged. In any two adjacent rows, the scanning direction of one row is from left to right, while the scanning direction of the other row is from right to left. Rows arranged from left to right are called forward rows, and rows arranged from right to left are called reverse rows. or Vertical raster scanning: Elements within a single compressed cell are arranged one by one along the vertical direction. After one column is completed, the next column is arranged. The scanning direction within all columns is either from top to bottom or from bottom to top. or Vertical back-and-forth scanning, also known as forward and backward scanning or bow scanning, involves arranging elements one by one along a vertical direction within a single compressed cell. After arranging one column, the next column is then arranged. In any two adjacent columns, one column is arranged from top to bottom while the other column is arranged from bottom to top. Columns arranged from top to bottom are called forward columns, and columns arranged from bottom to top are called reverse columns.

[0025] The first element in a string scan, i.e., the permutation, is called the starting element, and the last element in a string scan, i.e., the permutation, is called the ending element.

[0026] A special case of string prediction is block prediction, also known as block matching. In this special case, each string forms a rectangular block, or even a single string or block constitutes a complete compression unit. In this special case, encoding and decoding can proceed directly in blocks, rather than according to the form and order of element scanning. Alternatively, blocks and strings can be encoded and decoded in the same form and order as element scanning.

[0027] In this application, strings and blocks are collectively referred to as strings. Therefore, string prediction is also a general term for string prediction and block prediction.

[0028] A crucial factor affecting the coding efficiency of string prediction is the reference set, often referred to as the reference range, which is the range of possible positions of reference elements pointed to by the string vector. A larger reference range yields more candidate reference elements, increasing the likelihood of finding suitable ones and thus improving coding efficiency. However, a larger reference range also increases implementation complexity and cost. In particular, when the reference range becomes excessively large, reference elements cannot be stored internally in the codec chip and must be placed in a dedicated external memory chip. This not only increases cost but also severely bottlenecks read / write bandwidth and pixel processing throughput, potentially even hindering the encoding and decoding of high-resolution images and videos.

[0029] In string prediction techniques, to reduce implementation complexity and cost, the reference range is typically limited to a predetermined fixed size (i.e., the number of elements), such as being restricted to the largest compression unit or the nearest largest compression unit where the current compression unit is located. Even with a limited reference range, since its elements are all within the codec chip, read / write bandwidth and element processing throughput, especially the throughput of reading reference elements, remain bottlenecks. For example, if only a reference element capable of generating at most one reconstructed element can be read per clock cycle, the throughput of generating reconstructed elements will never exceed one element per clock cycle.

[0030] Point prediction is another special case of string prediction and an effective means of data compression.

[0031] Point prediction technology identifies the positions of data elements that have undergone predetermined encoding and decoding and whose values ​​frequently recur within or near the current compressed unit as "frequent positions" within the dataset. These positions are stored in a frequent position array or a subset thereof. Each frequent position in the array or subset is identified by an index, called a frequent position index or frequent position address. Data elements at frequent positions are used as reference elements, prediction elements, or matching elements. For a given string of equal values ​​to be encoded or decoded within a single compression unit, only an index parameter and a repetition count parameter for the frequently occurring position indicated by the index are needed. This indicates that the reference element for all elements of the equal string is an element at the frequently occurring position indicated by the index (this frequently occurring position can be the position of an element in the dataset preceding the equal string, or the position of an element within the equal string itself). Alternatively, it indicates that the value of all elements in the equal string is equal to the value of the element at the frequently occurring position indicated by the index (this frequently occurring position can be the position of an element in the dataset preceding the equal string, or the position of an element within the equal string itself). This eliminates the need to record the value of each element within the equal string individually, thus achieving data compression. The values ​​of elements in the equal string used for reconstruction are typically equal to or close to (i.e., the error does not exceed a predetermined threshold) the values ​​of the original elements, thereby ensuring high quality and low distortion of the reconstructed elements while achieving data compression. Frequently occurring positions are usually represented in the form of point vectors; therefore, the frequently occurring position index is also called the point vector address. The frequently occurring position array is usually a point vector array, that is, an array that stores the parameters of the point vectors. The parameters of the point vectors include at least the point vector itself and information related to the point vector. The point vector is usually represented by coordinates in a predetermined form, called the position coordinates of the point vector. The point vector array is sometimes also called the point prediction array or the point prediction information table.

[0032] For example, given a current string arranged according to a predetermined scanning method, if a point vector can be found within a persistent position array or a subset thereof, pointing to a reference element whose value is the predicted value of all elements in the current string, and using the predicted value as the value of the reconstructed element, then only the point vector address and the current string length are needed to record the value of each element in the current string. This eliminates the need to record the value of each element individually, thus completely representing all elements and their values ​​in the current string. The number of bits consumed in recording these two parameters is often far less than the number of bits consumed in recording the value of each element individually, thereby achieving data compression.

[0033] A persistent position element, if it appears in a positive position (meaning the position specified by the point vector at that persistent position is positive), has all N components, including the secondary component. The compressed data stream contains all N components, including the secondary component. The encoder writes the values ​​of all N components, including the secondary component, into the compressed data stream, while the decoder obtains the values ​​of all N components, including the secondary component, from the compressed data stream. Otherwise, it only has a primary component and no secondary components. The encoder only writes the values ​​of the primary components into the compressed data stream, and the decoder obtains only the values ​​of the primary components from the compressed data stream. The values ​​of the secondary components of a persistent position element are stored in the secondary component space at the positive position. A persistent position element serves as both a reconstructed element at its location and a reference element for other elements, used to obtain the reconstructed values ​​of other elements. In this case, the persistent position element is also called a persistent position reference element.

[0034] In an integer compression unit using point prediction encoding and decoding, there may be unmatched elements where no reference element can be found. If an unmatched element appears in a positive position, it has all N components, including the secondary component. The compressed data stream contains all N components, including the secondary component. The encoder writes the values ​​of all N components, including the secondary component, into the compressed data stream, and the decoder obtains the values ​​of all N components, including the secondary component, from the compressed data stream. Otherwise, it only has the primary component and no secondary component. The encoder only writes the value of the primary component into the compressed data stream, and the decoder obtains only the value of the primary component from the compressed data stream. Unmatched elements can also be used as reference elements for other elements to obtain their reconstructed values; in this case, the unmatched element is also called an unmatched reference element.

[0035] Point prediction technology divides the elements of a single compression unit into three types of string elements of variable length, along a scan path formed by continuously sorting them according to a predetermined scan method: String Type 1: Equivalent String. In an equivalent string, the reference element or predicted element of all elements is equal to the value of an element at a constant position. An equivalent string uses a constant position and its point vector index (also simply called the constant position index or point vector index) to represent the constant position and its point vector used by the equivalent string. This specifies which point vector points to the data element at the constant position whose value the reference element or predicted element of all elements in the equivalent string is equal to. String type 2: Unmatched element string. All elements in an unmatched element string are unmatched elements, i.e., unpredictable elements; an unmatched element string is also represented by a specific dot vector index value, indicating that the string is an unmatched element string rather than an equality string; String Type 3: Unit Basis Vector String. A unit basis vector string is a string that is located one unit distance directly above the current string, referring to the reference string. In horizontal scan mode, "directly above" means the directly adjacent string above, while in vertical scan mode, "directly above" means the directly adjacent string to the left. Therefore, a unit basis vector string is also called a copy of the string above.

[0036] As can be seen, both the equality string and the unmatched string are represented and distinguished using a dot vector index. Therefore, the equality string and the unmatched string are collectively referred to as dot vector indexed representation strings.

[0037] Dot vector index representations and unit basis vector strings are two types of strings with very different properties, requiring different encoding and decoding methods. Therefore, in dot prediction techniques, consecutively occurring dot vector index representations are treated as a group, and a variable called the string count and its syntax elements are used to indicate how many dot vector index representations are in the group. Based on the string count, the dot vector index representations in the group are encoded and decoded one by one.

[0038] Since the number of strings may be zero, existing point prediction techniques typically set the number of strings to be greater than or equal to zero. However, in reality, the number of strings is unlikely to be zero in most cases. In these cases, existing point prediction techniques still set the range of the number of strings to be greater than or equal to zero, which is larger than actually needed. The larger the range, the more bits are consumed in encoding and decoding the values ​​within the range using a truncated code with the maximum value, resulting in lower coding efficiency and severely reducing the efficiency of point prediction coding. Summary of the Invention

[0039] To address this problem in dot prediction coding, this invention provides a method and apparatus for encoding and decoding in which the number of strings in the first element of an integer compression unit and the number of strings in the remaining elements have different value ranges. When encoding and decoding a group of consecutive dot vector index representation strings, a variable called the string number and its syntax elements, representing the number of dot vector index representation strings in the group, are first encoded and decoded. The value range of the number of strings in the first element of the integer compression unit is greater than or equal to 0, and the value range of the number of strings after the first element is greater than or equal to 1. This reduces the value range of the number of strings after the first element, effectively improving the efficiency of dot prediction coding.

[0040] Preferably, The string number is encoded or decoded using one or a combination of the following codes: 1) Truncated binary code with the maximum value; 2) Truncated unary codes with the maximum value; 3) Truncated 0th-order exponent Golomb codes with maximum values; 4) Truncated first-order exponent Golomb codes with maximum values; 5) Truncated k-order exponential Golomb codes with maximum values; 6) The codeword consists of two parts: a prefix and an infix. The prefix is ​​a fixed-length code of length n-1, and the infix is ​​a fixed-length code of length 1 or empty. Here, n = Ceil(Log2(V)), Ceil() is the floor function, and Log2() is the logarithm function with base 2. Here, V is the maximum number of strings.

[0041] Preferably, For the first string count in the entire compression unit, the string count itself is encoded and decoded, and its value range is greater than or equal to 0. For the string counts after the first one, the string count minus one is encoded and decoded, and its value range is also greater than or equal to 0. The syntax description table describing the above characteristics includes at least: FirstStringNumFlag = 1 while (NumCodedPixel < NumTotalPixel) { Isc_evs_ups_num_minus1 / / Syntax element representing the number of strings if (FirstStringNumFlag) {get the string number StringNum from isc_evs_ups_num_minus1} StringNum = IscEvsUPSNumMinus1 FirstStringNumFlag = 0 } Else StringNum = IscEvsUPSNumMinus1 + 1 for (k = 0; k < StringNum; k++) { / / Encode and decode StringNum point vector indices representing strings one by one Isc_evs_ups_coding(x0,y0,width,height,component) } if (NumCodedPixel < NumTotalPixel) { Encoding and decoding a unit basis vector string } } In the above grammar description table, FirstStringNumFlag is a variable that indicates whether the current string count is the count of the first string or the count of the strings after the first string; FirstStringNumFlag of 1 indicates that the current string count is the count of the first string; FirstStringNumFlag of 0 indicates that the current string count is the count of the strings after the first string. The `while (NumCodedPixel < NumTotalPixel)` method means that the operation within the curly braces below is executed repeatedly while there are still elements in the entire compression unit that have not been fully encoded or decoded. isc_evs_ups_num_minus1 is a syntax element representing the number of strings, used to obtain the number of strings StringNum; the value range of isc_evs_ups_num_minus1 is greater than or equal to 0; for the first string number in an integer compression unit, isc_evs_ups_num_minus1 represents the string number itself, and for the string numbers after the first one, isc_evs_ups_num_minus1 represents the string number minus one; The `for (k = 0; k < StringNum; k++)` method executes the operation within the curly braces `StringNum` times, which means encoding and decoding the `StringNum` dot vector index representation strings one by one. isc_evs_ups_coding(x0,y0,width,height,component) is a function that encodes and decodes a string representing a dot vector index. The condition `if (NumCodedPixel < NumTotalPixel)` indicates that if there are still elements in the entire compression unit that have not been encoded or decoded, then the operation in the curly braces below is performed, which is to encode and decode a unit base vector string.

[0042] According to a first aspect of the present invention, a point prediction encoding method is provided, characterized by comprising at least the following steps: 1) Input the raw data of at least one complete compression unit. 2) At least the entire compression unit is subjected to point predictive coding. When encoding a set of consecutive point vector index representation strings, a variable called the string number, representing the number of point vector index representation strings in the set, and its syntax elements are first encoded. The value of the string number of the first string in the entire compression unit is greater than or equal to 0, and the value of the string number after the first string is greater than or equal to 1. 3) Output at least the compressed data bitstream of the entire compression unit.

[0043] According to a second aspect of the present invention, a point prediction coding apparatus is provided, characterized in that it comprises at least the following modules: 1) Raw data input module: Input the raw data of at least one complete compression unit. 2) Dot Predictive Coding Module: At least the entire compression unit undergoes dot predictive coding. When encoding a set of consecutive dot vector index representation strings, a variable called the string count, representing the number of dot vector index representation strings in the set, and its syntax elements are first encoded. The value of the first string count in the entire compression unit is greater than or equal to 0, and the value of the string counts after the first one is greater than or equal to 1. 3) Compressed data stream output module: outputs at least the compressed data stream of the entire compression unit.

[0044] According to a third aspect of the present invention, a point prediction decoding method is provided, characterized by comprising at least the following steps: 1) Input at least one full compression unit of compressed data bitstream. 2) At least for the integer compression unit, point prediction decoding is performed. When decoding a set of consecutive point vector index representation strings, a variable called the string number and its syntax elements, representing the number of point vector index representation strings in the set, are decoded first. The value of the string number in the first integer compression unit is greater than or equal to 0, and the value of the string number after the first one is greater than or equal to 1. 3) Output at least the reconstructed elements of the entire compression unit.

[0045] According to a fourth aspect of the present invention, a point prediction decoding apparatus is provided, characterized in that it comprises at least the following modules: 1) Compressed data stream input module: This module requires at least one complete compression unit of compressed data stream as input. 2) Point Prediction Decoding Module: At least for the entire compression unit, point prediction decoding is performed. When decoding a group of consecutive point vector index representation strings, a variable called the string count, representing the number of point vector index representation strings in the group, and its syntax elements are first decoded. The value of the string count for the first string in the entire compression unit is greater than or equal to 0, and the value of the string count for subsequent strings is greater than or equal to 1. 3) Reconstruction element output module: Outputs at least the reconstruction elements of the entire compression unit.

[0046] This invention is applicable to the encoding and decoding of data with lossy compression, and it is also applicable to the encoding and decoding of data with lossless compression. This invention is applicable to the encoding and decoding of one-dimensional data such as string data, byte string data, one-dimensional graphics, or fractal-dimensional graphics, and it is also applicable to the encoding and decoding of two-dimensional or higher-dimensional data such as images, image sequences, or video data.

[0047] In this invention, the data compression involves one or a combination of the following types of data. 1) One-dimensional data; 2) Two-dimensional data; 3) Multidimensional data; 4) Graphics; 5) Fractal dimensions; 6) Images; 7) Image sequences; 8) Video; 9) Audio; 10) Documents; 11) bytes; 12) bits; 13) Pixels; 14) Three-dimensional scene; 15) A sequence of continuously changing 3D scenes; 16) Virtual reality scenarios; 17) A sequence of continuously changing virtual reality scenes. 18) Images in pixel format; 19) Transform domain data of the image; 20) A collection of bytes with two or more dimensions; 21) A set of bits with two or more dimensions; 22) A set of pixels; 23) A set of single-component pixels; 24) A set of three-component pixels (R, G, B, A); 25) A set of three-component pixels (Y, U, V); 26) A set of three-component pixels (Y, Cb, Cr); 27) A set of three-component pixels (Y, Cg, Co); 28) A set of four-component pixels (C, M, Y, K); 29) A set of four-component pixels (R, G, B, A); 30) A set of four-component pixels (Y, U, V, A); 31) A set of four-component pixels (Y, Cb, Cr, A); 32) A set of four-component pixels (Y, Cg, Co, A).

[0048] The technical features of the present invention have been illustrated above through several specific examples. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. Attached Figure Description

[0049] Figure 1 This is a schematic diagram of the encoding method or apparatus of the present invention.

[0050] Figure 2 This is a schematic diagram of the decoding method or apparatus of the present invention. Detailed Implementation

[0051] The following are further implementation details of the present invention.

[0052] Example 1 In the encoding method, encoding device, or decoding method or decoding device, the original data is an array or sequence of two-dimensional data elements including images, image sequences, and videos. The compressed subsets involved in point prediction include images, sub-images of images, tiles, slices, and patches; The maximum compression units involved in point prediction include tiles, maximum coding units (LCUs), and coding tree units (CTUs). The entire compression unit includes a macroblock, a coding unit CU, a sub-region of CU, a sub-coding unit SubCU, a prediction block, a prediction unit PU, a sub-region of PU, a sub-prediction unit SubPU, a transform block, a transform unit TU, a sub-region of TU, ​​and a sub-transform unit SubTU.

[0053] Example 2 In the encoding method or encoding device or decoding method or decoding device The string number is encoded or decoded using one or a combination of the following codes: 1) Truncated binary code with the maximum value; 2) Truncated unary codes with the maximum value; 3) Truncated 0th-order exponent Golomb codes with maximum values; 4) Truncated first-order exponent Golomb codes with maximum values; 5) Truncated k-order exponential Golomb codes with maximum values; 6) The codeword consists of two parts: a prefix and an infix. The prefix is ​​a fixed-length code of length n-1, and the infix is ​​a fixed-length code of length 1 or empty. Here, n = Ceil(Log2(V)), Ceil() is the floor function, and Log2() is the logarithm function with base 2. Here, V is the maximum number of strings.

[0054] Example 3 In the encoding method or encoding device or decoding method or decoding device For the first string count in the entire compression unit, the string count itself is encoded and decoded, and its value range is greater than or equal to 0. For the string counts after the first one, the string count minus one is encoded and decoded, and its value range is also greater than or equal to 0. The syntax description table describing the above characteristics includes at least: FirstStringNumFlag = 1 while (NumCodedPixel < NumTotalPixel) { Isc_evs_ups_num_minus1 / / Syntax element representing the number of strings if (FirstStringNumFlag) {get the string number StringNum from isc_evs_ups_num_minus1} StringNum = IscEvsUPSNumMinus1 FirstStringNumFlag = 0 } else StringNum = IscEvsUPSNumMinus1 + 1 for (k = 0; k < StringNum; k++) { / / Encode and decode StringNum point vector indices representing strings one by one isc_evs_ups_coding(x0,y0,width,height,component) } if (NumCodedPixel < NumTotalPixel) { Encoding and decoding a unit basis vector string } } In the above grammar description table, FirstStringNumFlag is a variable that indicates whether the current string count is the count of the first string or the count of the strings after the first string; FirstStringNumFlag of 1 indicates that the current string count is the count of the first string; FirstStringNumFlag of 0 indicates that the current string count is the count of the strings after the first string. The `while (NumCodedPixel < NumTotalPixel)` method means that the operation within the curly braces below is executed repeatedly while there are still elements in the entire compression unit that have not been fully encoded or decoded. isc_evs_ups_num_minus1 is a syntax element representing the number of strings, used to obtain the number of strings StringNum; the value range of isc_evs_ups_num_minus1 is greater than or equal to 0; for the first string number in an integer compression unit, isc_evs_ups_num_minus1 represents the string number itself, and for the string numbers after the first one, isc_evs_ups_num_minus1 represents the string number minus one; The `for (k = 0; k < StringNum; k++)` method executes the operation within the curly braces `StringNum` times, which means encoding and decoding the `StringNum` dot vector index representation strings one by one. isc_evs_ups_coding(x0,y0,width,height,component) is a function that encodes and decodes a string representing a dot vector index. The condition `if (NumCodedPixel < NumTotalPixel)` indicates that if there are still elements in the entire compression unit that have not been encoded or decoded, then the operation in the curly braces below is performed, which is to encode and decode a unit base vector string.

Claims

1. A method for decoding a string of consecutively occurring point vector index representations, characterized in that... It should include at least the following steps: 1) Input at least one encoded unit of compressed data bitstream. 2) At least for the aforementioned coding unit, point prediction decoding is performed. When decoding a set of consecutively occurring point vector index representation strings, a variable called the string number, representing the number of point vector index representation strings in the set, and its syntax elements are first decoded. The value of the first string number in the coding unit is greater than or equal to 0, and the value of the string number after the first one is greater than or equal to 1. 3) Output at least the reconstructed pixels of the encoding unit; It also has the following characteristics: For the first string count in the encoding unit, the string count itself is decoded, and its value range is greater than or equal to 0. For the string counts after the first one, the string count is reduced by one and then decoded, so its value range also becomes greater than or equal to 0. The grammatical description table describing the above features should include at least: FirstStringNumFlag = 1 while (NumCodedPixel < NumTotalPixel) { isc_evs_ups_num_minus1 / / Syntax element representing the number of strings if (FirstStringNumFlag) {get the string number StringNum from isc_evs_ups_num_minus1} StringNum = IscEvsUPSNumMinus1 FirstStringNumFlag = 0 } else StringNum = IscEvsUPSNumMinus1 + 1 for (k = 0; k < StringNum; k++) { / / Decode each StringNum dot vector index representing the string one by one isc_evs_ups_coding(x0,y0,width,height,component) } if (NumCodedPixel < NumTotalPixel) { Decoding a unit basis vector string } } In the above grammar description table, FirstStringNumFlag is a variable that indicates whether the current string count is the count of the first string or the count of the strings after the first string; FirstStringNumFlag of 1 indicates that the current string count is the count of the first string; FirstStringNumFlag of 0 indicates that the current string count is the count of the strings after the first string. The `while (NumCodedPixel < NumTotalPixel)` method means that the operation within the curly braces below is executed repeatedly while there are still pixels in the encoding unit that have not been fully decoded. isc_evs_ups_num_minus1 is a syntax element representing the number of strings, used to obtain the number of strings StringNum; the value range of isc_evs_ups_num_minus1 is greater than or equal to 0; for the first string number in the encoding unit, isc_evs_ups_num_minus1 represents the string number itself, and for the string numbers after the first one, isc_evs_ups_num_minus1 represents the string number minus one; The `for (k = 0; k < StringNum; k++)` method means executing the operation within the curly braces `StringNum` times, which means decoding the `StringNum` dot vector index representation strings one by one. isc_evs_ups_coding(x0,y0,width,height,component) is a function that decodes a dot vector index representation of a string; The condition `if (NumCodedPixel < NumTotalPixel)` means that if there are still pixels in the encoding unit that have not been decoded, then the operation in the curly braces below is performed, which is to decode a unit basis vector string.

2. A decoding device for a string of consecutively occurring point vector index representations, characterized in that... It should include at least the following modules: 1) Compressed data stream input module: This module requires at least one encoded unit of compressed data stream as input. 2) Dot Prediction Decoding Module: At least the encoding unit undergoes dot prediction decoding. When decoding a group of consecutive dot vector index representation strings, a variable called the string count, representing the number of dot vector index representation strings in the group, and its syntax elements are first decoded. The value of the first string count in the encoding unit is greater than or equal to 0, and the value of the string counts after the first one is greater than or equal to 1. 3) Reconstructed pixel output module: outputs at least the reconstructed pixels of the encoding unit; It also has the following characteristics: For the first string count in the encoding unit, the string count itself is decoded, and its value range is greater than or equal to 0. For the string counts after the first one, the string count is reduced by one and then decoded, so its value range also becomes greater than or equal to 0. The grammatical description table describing the above features should include at least: FirstStringNumFlag = 1 while (NumCodedPixel < NumTotalPixel) { isc_evs_ups_num_minus1 / / Syntax element representing the number of strings if (FirstStringNumFlag) {get the string number StringNum from isc_evs_ups_num_minus1} StringNum = IscEvsUPSNumMinus1 FirstStringNumFlag = 0 } else StringNum = IscEvsUPSNumMinus1 + 1 for (k = 0; k < StringNum; k++) { / / Decode each StringNum dot vector index representing the string one by one isc_evs_ups_coding(x0,y0,width,height,component) } if (NumCodedPixel < NumTotalPixel) { Decoding a unit basis vector string } } In the above grammar description table, FirstStringNumFlag is a variable that indicates whether the current string count is the count of the first string or the count of the strings after the first string; FirstStringNumFlag of 1 indicates that the current string count is the count of the first string; FirstStringNumFlag of 0 indicates that the current string count is the count of the strings after the first string. The `while (NumCodedPixel < NumTotalPixel)` method means that the operation within the curly braces below is executed repeatedly while there are still pixels in the encoding unit that have not been fully decoded. isc_evs_ups_num_minus1 is a syntax element representing the number of strings, used to obtain the number of strings StringNum; the value range of isc_evs_ups_num_minus1 is greater than or equal to 0; for the first string number in the encoding unit, isc_evs_ups_num_minus1 represents the string number itself, and for the string numbers after the first one, isc_evs_ups_num_minus1 represents the string number minus one; The `for (k = 0; k < StringNum; k++)` method means executing the operation within the curly braces `StringNum` times, which means decoding the `StringNum` dot vector index representation strings one by one. isc_evs_ups_coding(x0,y0,width,height,component) is a function that decodes a dot vector index representation of a string; The condition `if (NumCodedPixel < NumTotalPixel)` means that if there are still pixels in the encoding unit that have not been decoded, then the operation in the curly braces below is performed, which is to decode a unit basis vector string.