Image reconstruction method and device, image compression method and device

Abstract

The embodiment of the present invention provides an image reconstruction method and device, and an image compression method and device, which belong to the field of image processing technology. The image reconstruction method comprises: for the measurement value of the original image, repeatedly performing the following steps to perform iterative calculation until a predetermined number of iterations K is reached, and taking the first measurement value #imgabs0# updated for the Kth time as the reconstructed image x: interpolating and enlarging the first measurement value y1 to the same size as the original image to obtain the interpolated image y f ; For the interpolated image y f Perform Huber transform and compressed sensing transform to obtain the Huber transformed image y H and the image y after compressed sensing transformation fC ; Calculate the image y after compressed sensing transformation fC and the second measured value y2, and performing a compressed sensing inverse transform on the difference to obtain a residual r; and using the residual r and the Huber transformed image y H , the first measurement value y1 is updated. The reconstructed image has a higher peak signal-to-noise ratio and a better visual effect.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to an image reconstruction method and device, and an image compression method and device. Background Art

[0002] Deep space exploration can obtain information about unknown areas, and deep space images taken by CCD cameras or astronomical telescopes are the most intuitive representation of information. However, due to the contradiction between massive data and relatively narrow channel bandwidth, images need to be compressed on-orbit to reduce the difficulty of deep space communication. The most widely used image compression method in deep space exploration is JPEG / JPEG 2000. JPEG is a lossy compression algorithm based on quantization after discrete cosine transform. It has good performance at medium and high bit rates and low distortion, but has high distortion at low bit rates.

[0003] To solve the above problems, a deep image compression algorithm based on region of interest is proposed. The algorithm demarcates the region of interest through target recognition method, extracts more coefficients for the frequency transform of the region of interest, and extracts a small number of coefficients for the non-region of interest, so as to obtain better target images at low bit rate. However, the accuracy of target recognition has an impact on the results. If the accuracy is not good, important targets may be missed. Summary of the invention

[0004] The purpose of the embodiments of the present invention is to provide an image reconstruction method and device to solve the problem of high distortion rate at low bit rate existing in the existing image compression technology (especially astronomical image compression technology).

[0005] In order to achieve the above-mentioned object, an embodiment of the present invention provides an image reconstruction method, the method comprising: for the measurement value of the original image, repeatedly performing the following steps to perform iterative calculations until a predetermined number of iterations K is reached, and then the first measurement value after the Kth update is As the reconstructed image x, wherein the measurement values ​​include the first measurement value y 1 and the second measurement y 2 , wherein the first measurement value y 1 The second measurement value y is obtained by downsampling the original image. 2 The first measurement value y is obtained by performing a compressed sensing transformation on the original image: 1 Interpolate and enlarge to the same size as the original image to obtain the interpolated image y f ; For the interpolated image y f Perform Huber transform to obtain the image y after Huber transform H ; For the interpolated image y fPerform the compressed sensing transformation to obtain the image y after the compressed sensing transformation fC ; Calculate the image y after the compressed sensing transformation fC With the second measured value y 2 and performing a compressed sensing inverse transform on the difference to obtain a residual r; and using the residual r and the Huber transformed image y H , for the first measured value y 1 to update.

[0006] Optionally, the Huber transformed image y is obtained according to the following Huber function: H :

[0007]

[0008] Wherein, α is the preset Huber transform threshold.

[0009] Optionally, the first measured value y 1 The updating includes using the gradient descent method to update the first measurement value y 1 Update, including: according to the residual r, the image y after Huber transformation H , calculate the gradient grad; and use the gradient to calculate the first measurement value y 1 to update.

[0010] Optionally, the gradient grad is calculated according to the following formula:

[0011] grad=y H +λr

[0012] And / or the first measured value y according to the following formula 1 To update:

[0013] y 1 =y 1 -grad·s

[0014] Wherein λ is a predetermined weight coefficient, and s is a predetermined step length.

[0015] Optionally, the compressed sensing transform is a noiselet transform, and the compressed sensing inverse transform is an inverse noiselet transform.

[0016] Correspondingly, an embodiment of the present invention also provides an image compression method, which includes: using a constructed joint perception matrix Φ to compress an original image to obtain a measurement value of the original image, wherein the joint perception matrix Φ includes a downsampling matrix and a compressed sensing transformation matrix.

[0017] Optionally, the compressed sensing transformation matrix is ​​a noiselet transformation matrix.

[0018] Correspondingly, an embodiment of the present invention further provides an image reconstruction device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor executes the computer program to implement the above-mentioned image reconstruction method.

[0019] Correspondingly, an embodiment of the present invention further provides an image compression device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor executes the computer program to implement the above-mentioned image compression method.

[0020] Correspondingly, an embodiment of the present invention further provides a machine-readable storage medium, on which instructions are stored, and the instructions are used to enable a machine to execute the above-mentioned image reconstruction method and / or the above-mentioned image compression method.

[0021] The image compression method provided by the embodiment of the present invention obtains image compression data through a joint perception matrix, thereby greatly reducing the amount of transmitted data. Based on the image reconstruction method provided by the embodiment of the present invention, for the image compression data obtained through the joint perception matrix, a reconstructed image with a higher peak signal-to-noise ratio than the traditional JPEG image can be obtained, and a better visual effect can be obtained. The image compression method and the image reconstruction method can be applied to the field of astronomical images, and the amount of data of astronomical image compression data can be greatly reduced, and the reconstructed image has a higher peak signal-to-noise ratio than the traditional JPEG image and has a better visual effect. In addition, the image compression method and the image reconstruction method can also be applied to the transmission of massive data of large array detectors.

[0022] Other features and advantages of the embodiments of the present invention will be described in detail in the subsequent detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following specific implementations, they are used to explain the embodiments of the present invention, but do not constitute a limitation on the embodiments of the present invention. In the accompanying drawings:

[0024] Figure 1 A schematic flow chart of an image reconstruction method according to an embodiment of the present invention is shown.

[0025] Figure 2 The Huber function curves at different α are shown.

[0026] Figure 3 A raw nebula image is shown.

[0027] Figure 4 Shown for Figure 3 The original nebula image and the reconstructed nebula image are shown.

[0028] Figure 5 A structural block diagram of an electronic device provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0029] The specific implementation of the embodiment of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the embodiment of the present invention, and is not used to limit the embodiment of the present invention.

[0030] Compression algorithms based on compressed sensing have been proposed for astronomical image compression. Compressed sensing (CS) uses the unique sparsity of natural signals and minimal sampling redundancy to reconstruct high-quality signals while significantly reducing the sampling rate. Compared with the traditional signal processing method in which the sampling process must comply with Shannon's sampling theorem, the compressed sensing method transfers the amount of computation from the encoding end of signal acquisition to the decoding end, which has significant advantages in resource-limited imaging technologies such as remote sensing, and is therefore also suitable for the acquisition and compression of astronomical images that are inherently sparse.

[0031] Figure 1 FIG. 4 shows a flow chart of an image reconstruction method according to an embodiment of the present invention. Figure 1 As shown, an embodiment of the present invention provides an image reconstruction method, which may include repeatedly performing steps S01-S05 for iterative calculations on the measurement value y of the original image until a predetermined number of iterations K is reached, and then performing the Kth updated first measurement value As the reconstructed image x, wherein the measurement values ​​include the first measurement value y 1 and the second measurement y 2 , wherein the first measurement value y 1 The second measurement value y is obtained by downsampling the original image. 2 Obtained by performing a compressed sensing transformation on the original image. K is a positive integer, indicating the number of iterations required, and can be set to any appropriate value as needed. The more iterations there are, the higher the quality of the reconstructed image obtained, and the time cost required will also increase accordingly. Therefore, when setting the K value, it is necessary to balance the image quality and time cost.

[0032] The measurement value of the original image is the data after the original image is compressed. The measurement value can be obtained by constructing a joint perception matrix Φ The joint sensing matrix includes the downsampling matrix and the compressed sensing transformation matrix. The measurement value y obtained through the joint sensing matrix Φ is expressed as:

[0033]

[0034] Among them, Φ C is the compressed sensing transformation matrix, Φ D is the downsampling matrix, f is the original image, y is the measured value, y 1 =Φ D f,y 2 =Φ C f.

[0035] The compressed sensing transformation matrix Φ C Preferably, the noiselet transformation matrix Φ N , so as to facilitate the acquisition of the measurement value y through hardware. However, the embodiment of the present invention is not limited to this. The compressed sensing transformation matrix Φ C It can also be a Bernoulli matrix, a Hadamard matrix, etc.

[0036] S01, for the first measured value y 1 Interpolate and enlarge to the same size as the original image to obtain the interpolated image y f .

[0037] Any suitable difference method may be used to perform step S03.

[0038] S02, for the interpolated image y f Perform Huber transform to obtain the image y after Huber transform H .

[0039] Specifically, the following Huber function can be used to perform the Huber transformation to obtain the Huber transformed image y H :

[0040]

[0041] Where α is the preset Huber transform threshold. The Huber function curves under different α are as follows: Figure 2 shown.

[0042] In the embodiment of the present invention, the Huber function is not limited to the above formula, and any other suitable Huber function may be used to perform Huber transformation.

[0043] S03, for the interpolated image y f Perform the compressed sensing transformation to obtain the image y after the compressed sensing transformation fC .

[0044] The compressed sensing transformation matrix used in this step is the same as the second measurement value y2 The same compressed sensing transformation matrix is ​​used.

[0045] S04, calculating the image y after the compressed sensing transformation fC With the second measured value y 2 The difference between them is calculated, and the compressed sensing inverse transform is performed on the difference to obtain the residual r.

[0046] The compressed sensing inverse transform corresponds to the compressed sensing transform. If the compressed sensing transform is the noiselet transform, then the compressed sensing inverse transform is the noiselet inverse transform.

[0047] S05, using the residual r and the image y after Huber transformation H , for the first measured value y 1 to update.

[0048] Any iterative solution method can be used to calculate the first measured value y 1 Perform updates, such as gradient descent or L1 norm.

[0049] Preferably, the first measurement value y can be calculated using a gradient descent method. 1 to update.

[0050] First, according to the residual r, the image y after Huber transformation H , calculate the gradient grad. Specifically, the gradient grad can be calculated according to the following formula:

[0051] grad=y H +λr (3)

[0052] Wherein λ is a predetermined weight coefficient, which can be set to any appropriate value as required.

[0053] Then use the gradient to calculate the first measured value y 1 Specifically, the first measurement value y can be updated according to the following formula 1 To update:

[0054] y 1 =y 1 -grad·s (4)

[0055] Wherein s is a predetermined step length, which can be set to any appropriate value as required, for example, can be set to 1.

[0056] Use the updated first measurement value, return to execute steps S01-S05, until a predetermined number of iterations K is reached, and then update the first measurement value after the Kth update As the reconstructed image x.

[0057] Correspondingly, an embodiment of the present invention also provides an image compression method, which includes using a constructed joint perception matrix Φ to compress an original image to obtain a measurement value of the original image, wherein the joint perception matrix Φ includes a downsampling matrix and a compressed sensing transformation matrix.

[0058] The measured values ​​accordingly include a first measured value y 1 and the second measurement y 2 , wherein the first measurement value y 1 The second measurement value y is obtained by downsampling the original image. 2 It is obtained by performing compressed sensing transformation on the original image.

[0059] The joint perception matrix can be constructed as Among them, Φ C is the compressed sensing transformation matrix, Φ D is the downsampling matrix. The measured value Then y 1 =Φ D f,y 2 =Φ C f, where f is the original image and y is the measured value.

[0060] The compressed sensing transformation matrix Φ C Preferably, the noiselet transformation matrix Φ N , so as to facilitate the acquisition of the measurement value y through hardware. However, the embodiment of the present invention is not limited to this. The compressed sensing transformation matrix Φ C It can also be a Bernoulli matrix, a Hadamard matrix, etc.

[0061] The following is Figure 3 The method provided by the present invention is further described by taking the original nebula image shown in FIG. 1 as an example of compression and reconstruction. In this embodiment, the original nebula image is 512×512, and the parameters α=0.01, λ=0.01, s=1, and K=50. The compressed sensing transform used is the noiselet transform, and the compressed sensing inverse transform is the noiselet inverse transform. Specifically, the following steps are included.

[0062] S11, the original nebula image is captured and then sensed using a joint sensing matrix to obtain the measured value The first measurement y 1 The value is obtained by downsampling the original image, and the second measurement value y 2Obtained by performing noiselet transformation on the original image. In this embodiment, the compression is 8 times, the size of the original nebula image after downsampling is 128×128, and the number of random sampling of noiselet transformation is 16384. The measured value is transmitted to the image reconstruction device, and the image reconstruction device performs image reconstruction based on the measured value. The image reconstruction device repeatedly performs the following steps to perform iterative calculations until the predetermined number of iterations is 50 times.

[0063] S12, for the first measured value y 1 Interpolate and enlarge to the same size as the original image to obtain the interpolated image y f .

[0064] S13, using formula (2) to calculate the interpolated image y f Perform Huber transform to obtain the image y after Huber transform H .

[0065] S14, for the interpolated image y f Perform noiselet transformation to obtain the image y after noiselet transformation fN That is, y fN =Φ N y f , where Φ N is the noiselet transformation matrix.

[0066] S15, calculating the image y after the noiselet transformation fN With the second measured value y 2 The difference between them is obtained by performing an inverse noiselet transform on the difference to obtain the residual r. That is, in It is the noiselet inverse transformation matrix.

[0067] S16, using the gradient descent method to calculate the first measurement value y 1 First, use formula (3) to calculate the gradient grad, and then use formula (4) to update the first measurement value y 1 to update.

[0068] Use the updated first measurement value, return to execute steps S12-S16, until the predetermined number of iterations reaches 50, and then update the first measurement value after the 50th update As the reconstructed image x. The reconstructed image x is Figure 4 shown.

[0069] The image compression method provided by the embodiment of the present invention obtains image compression data through a joint perception matrix, thereby greatly reducing the amount of transmitted data. Based on the image reconstruction method provided by the embodiment of the present invention, for the image compression data obtained through the joint perception matrix, a reconstructed image with a higher peak signal-to-noise ratio than the traditional JPEG image can be obtained, and a better visual effect can be obtained. The image compression method and the image reconstruction method can be applied to the field of astronomical images, and the amount of data of astronomical image compression data can be greatly reduced, and the reconstructed image has a higher peak signal-to-noise ratio than the traditional JPEG image and has a better visual effect. In addition, the image compression method and the image reconstruction method can also be applied to the transmission of massive data of large array detectors.

[0070] An embodiment of the present invention further provides an image reconstruction device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the image reconstruction method according to any embodiment.

[0071] An embodiment of the present invention further provides an image compression device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the image compression method according to any embodiment.

[0072] The processor includes a kernel, and the kernel retrieves the corresponding program unit from the memory. One or more kernels can be provided, and the image reconstruction method and / or image compression method described in any embodiment of the present invention can be executed by adjusting kernel parameters.

[0073] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0074] An embodiment of the present invention provides a machine-readable storage medium, on which instructions are stored, and the instructions are used to enable a machine to execute the image reconstruction method and / or image compression method according to any embodiment of the present invention.

[0075] An embodiment of the present invention provides an electronic device, such as Figure 5As shown, the electronic device 50 includes at least one processor 501, at least one memory 502 and a bus 503 connected to the processor 501; wherein the processor 501 and the memory 502 communicate with each other through the bus 503; the processor 501 is used to call the program instructions in the memory 502 to execute the image reconstruction method according to any embodiment of the present invention. The electronic device of the embodiment of the present invention can be a server, a PC, a PAD, a mobile phone, etc.

[0076] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program initialized with the steps of the image reconstruction method according to any embodiment of the present invention.

[0077] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0078] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0079] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0080] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0081] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0082] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0083] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0084] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0085] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. An image reconstruction method, characterized in that: The method comprises the following steps of repeatedly performing iterative operations on the measurement value of the original image until a predetermined number of iterations K is reached, and then the first measurement value after the Kth update is calculated. As a reconstructed image x, wherein the measurement value includes a first measurement value y1 and a second measurement value y2, wherein the first measurement value y1 is obtained by downsampling the original image, and the second measurement value y2 is obtained by performing a compressed sensing transform on the original image: The first measurement value y1 is interpolated and enlarged to the same size as the original image to obtain the interpolated image y f ; For the interpolated image y f Perform Huber transform to obtain the image y after Huber transform H ; For the interpolated image y f Perform the compressed sensing transformation to obtain the image y after the compressed sensing transformation fC ; Calculate the image y after the compressed sensing transformation fC and the second measurement value y2, and performing a compressed sensing inverse transform on the difference to obtain a residual r; According to the residual r, the image y after Huber transformation H , calculate the gradient grad; and Using the gradient to update the first measurement value y1, The gradient grad is calculated according to the following formula: grad=y H +λr The first measured value y1 is updated according to the following formula: y1=y1-grad·s Wherein λ is a predetermined weight coefficient, and s is a predetermined step length.

2. The method according to claim 1, characterized in that The Huber transformed image y is obtained according to the following Huber function H : Wherein, α is the preset Huber transform threshold.

3. The method according to claim 1, characterized in that The compressed sensing transform is a noiselet transform, and the compressed sensing inverse transform is an inverse noiselet transform.

4. An image reconstruction device, characterized in that: The device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the image reconstruction method according to any one of claims 1 to 3.

5. A machine-readable storage medium, characterized in that: The machine-readable storage medium stores instructions for causing a machine to execute the image reconstruction method according to any one of claims 1 to 3.

Images