Image transmission method, device, equipment, medium and computer program product

Abstract

The invention provides an image transmission method, device and equipment, a medium and a computer program product. The method comprises the following steps: acquiring a feedback signal-to-noise ratio of an image transmission channel; coding a target image based on the feedback signal-to-noise ratio to obtain an image coding vector; determining image coding and decoding loss based on the target image and the decoded image; the decoded image is obtained by decoding the image coding vector; and under the condition that a to-be-processed image is transmitted through the image transmission channel, performing coding and decoding processing on the to-be-processed image based on the image coding and decoding loss to obtain an image transmission result. According to the invention, the encoding and decoding process of the image is dynamically adjusted through the signal-to-noise ratio of the channel, and the accuracy of image encoding and decoding in the image transmission process is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular, to an image transmission method, apparatus, device, medium, and computer program product. Background Art

[0002] Regarding the encoding and decoding problems in the process of image transmission, existing solutions use deep learning to explore the rules of image data and adjust the image encoding and decoding process to resist the losses caused by transmission channel noise, that is, to reduce the losses in the process of image transmission. However, the losses in the process of image transmission are also related to the transmission channel itself. Ignoring the influence of the changes in the transmission channel itself on the transmission losses will make the adjustment of the image encoding and decoding process inapplicable to most image data transmission processes. Therefore, how to dynamically adjust the image encoding and decoding through the channel signal-to-noise ratio has become an urgent technical problem to be solved. Summary of the Invention

[0003] The present invention provides an image transmission method, apparatus, device, medium, and computer program product to solve the defects of poor applicability and accuracy in the adjustment process of image encoding and decoding in the existing image transmission process, and improve the accuracy of image encoding and decoding in the process of image transmission.

[0004] The present invention provides an image transmission method, including the following steps: Obtain the feedback signal-to-noise ratio of the image transmission channel; Encode the target image based on the feedback signal-to-noise ratio to obtain an image encoding vector; Determine the image encoding and decoding loss based on the target image and the decoded image; the decoded image is obtained by decoding the image encoding vector; When transmitting the image to be processed through the image transmission channel, perform encoding and decoding processing on the image to be processed based on the image encoding and decoding loss to obtain an image transmission result.

[0005] According to the image transmission method provided by the present invention, the encoding the target image based on the feedback signal-to-noise ratio to obtain an image encoding vector includes: Determine the input image vector of the target image; the input image vector is determined based on the size of the target image; Obtain the encoding parameters of the joint channel encoder; the joint channel encoder is used to encode the target image; Encode the target image based on the feedback signal-to-noise ratio, the input image vector, and the encoding parameters to obtain an image encoding vector.

[0006] According to the image transmission method provided by the present invention, the image transmission method further includes: Determine the transfer function of the image transmission channel; Determine the vector to be decoded based on the transfer function, Gaussian distribution samples, and the image coding vector; the Gaussian distribution samples are vector samples conforming to the Gaussian distribution; Determine the mapping relationship between the vector to be decoded and the feedback signal-to-noise ratio; Decode the vector to be decoded based on the feedback signal-to-noise ratio, the mapping relationship, and decoding parameters to obtain the decoded image.

[0007] According to an image transmission method provided by the present invention, the determining the image coding and decoding loss based on the target image and the decoded image includes: Perform pixel comparison on the target image and the decoded image to obtain an image pixel loss; Determine an image discrimination loss based on a first determination probability and a second determination probability; the first determination probability is the probability obtained by the target discriminator determining the target image; the second determination probability is the probability obtained by the target discriminator determining the decoded image; Determine an image adversarial loss based on a third determination probability and the target image; the third determination probability is the probability obtained by the updated discriminator determining the decoded image; the updated discriminator is obtained by updating the target discriminator based on the first determination probability and the second determination probability; Determine an image smoothing loss based on the decoded image.

[0008] According to an image transmission method provided by the present invention, after the determining the image coding and decoding loss based on the target image and the decoded image, it includes: Update the target coding and decoding parameters based on at least one of a first update method, a second update method, and a third update method; the first update method is determined based on the image pixel loss; the second update method is determined based on the image adversarial loss and the image smoothing loss; the third update method is determined based on the image pixel loss, the image adversarial loss, and the image smoothing loss.

[0009] According to an image transmission method provided by the present invention, when transmitting the image to be processed through the image transmission channel, performing coding and decoding processing on the image to be processed based on the image coding and decoding loss to obtain an image transmission result includes: When transmitting the image to be processed through the image transmission channel, perform coding and decoding processing on the image to be processed based on the updated coding and decoding parameters to obtain an image transmission result; the updated coding and decoding parameters are obtained by updating the target coding and decoding parameters based on the update method corresponding to the feedback signal-to-noise ratio.

[0010] The present invention also provides an image transmission device, including the following modules: A feedback signal-to-noise ratio acquisition module, configured to acquire the feedback signal-to-noise ratio of an image transmission channel; An image coding vector determination module, configured to encode a target image based on the feedback signal-to-noise ratio to obtain an image coding vector; An image coding and decoding loss determination module, configured to determine an image coding and decoding loss based on the target image and a decoded image; the decoded image is obtained by decoding the image coding vector; An image transmission module, configured to, when transmitting an image to be processed through the image transmission channel, perform coding and decoding processing on the image to be processed based on the image coding and decoding loss to obtain an image transmission result.

[0011] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and running on the processor. When the processor executes the computer program, the image transmission method described in any one of the above is implemented.

[0012] The present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the image transmission method described in any one of the above is implemented.

[0013] The present invention also provides a computer program product, including a computer program. When the computer program is executed by a processor, the image transmission method described in any one of the above is implemented.

[0014] The image transmission method, device, equipment, medium, and computer program product provided by the present invention encode a target image based on the feedback signal-to-noise ratio of an image transmission channel to obtain an image coding vector; then, after decoding the image coding vector to obtain a decoded image, an image coding and decoding loss is determined based on the target image and the decoded image. Finally, when transmitting an image to be processed through the image transmission channel, coding and decoding processing is performed on the image to be processed based on the image coding and decoding loss to obtain an image transmission result. The present application dynamically adjusts the coding and decoding process of an image through the channel signal-to-noise ratio, improving the accuracy of image coding and decoding during the image transmission process. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0016] Figure 1It is one of the schematic flowcharts of the image transmission method provided by the present invention.

[0017] Figure 2 It is the second of the schematic flowcharts of the image transmission method provided by the present invention.

[0018] Figure 3 It is the schematic structural diagram of the image transmission device provided by the present invention.

[0019] Figure 4 It is the schematic structural diagram of the electronic device provided by the present invention. Detailed implementation manners

[0020] To make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions in the present invention will be clearly and completely described below with reference to the accompanying drawings in the present invention. Apparently, the described embodiments are some but not all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without making creative efforts based on the embodiments in the present invention belong to the scope of protection of the present invention.

[0021] The following combines Figures 1-4 to describe the image transmission method, device, equipment, medium and computer program product of the present invention.

[0022] Figure 1 It is one of the schematic flowcharts of the image transmission method provided by the present invention. As Figure 1 shown, the method includes the following: Step 100: Obtain the feedback signal-to-noise ratio of the image transmission channel; Specifically, for point-to-point image transmission with signal-to-noise ratio feedback, the signal-to-noise ratio (Signal to Interference plus Noise Ratio, SNR) of the image transmission channel can be determined on the joint source-channel encoder (for encoding the image) and the joint source-channel decoder (for decoding the image). The purpose of obtaining the channel signal-to-noise ratio is to fuse the channel signal-to-noise ratio with the image features, and then adjust the encoding and decoding process of the image based on the channel signal-to-noise ratio.

[0023] Step 200: Encode the target image based on the feedback signal-to-noise ratio to obtain an image encoding vector; Specifically, taking a target image with a size of H (height) × W (width) × C (channels) as an example, the target image is represented by a vector For the convenience of calculation, the feedback signal-to-noise ratio is represented by an integer The encoding function is , and a vector of real-valued channel input symbols is derived. The process of encoding the target image based on the feedback signal-to-noise ratio is shown in Formula 1, where is the size of the channel input symbols; is the parameter set of the joint source-channel encoder; is the set of real numbers. The encoder maps the n-dimensional vector of the integer-valued image to the real-valued channel input samples of dimensional vector.

[0024] ; (1) The encoded symbol is transmitted through a noisy channel represented by the function . Among them, the encoded symbol (channel input samples) is the image coding vector in this embodiment.

[0025] Step 300, determine the image coding and decoding loss based on the target image and the decoded image; the decoded image is obtained by decoding the image coding vector; Specifically, the representation of the channel output symbol received by the joint source-channel decoder is shown in Equation 2. Among them, the vector consists of independently and identically distributed samples, and each sample follows the Gaussian distribution , is the noise power.

[0026] ; (2) The joint channel decoder maps , and as shown in Equation 3 through the function is the set of integers; is the estimated value of the target image ; is the parameter set of the joint source-channel decoder; the distortion between the target image and the reconstructed image is shown in Equation 4, where the reconstructed image is the decoded image in this embodiment; the distortion is the image coding and decoding loss in this embodiment; and are the numerical values of the color components of each corresponding pixel.

[0027] ; (3) ; (4) Step 400, in the case of transmitting the image to be processed through the image transmission channel, perform coding and decoding processing on the image to be processed based on the image coding and decoding loss to obtain an image transmission result.

[0028] Specifically, the parameters of the codec are updated by gradient descent using the image codec loss to obtain the optimal encoder parameters and decoder parameters. As shown in Equation 5, where and are the optimal encoder parameters and the optimal decoder parameters respectively; is the joint probability distribution of the target image and the decoded image; is the probability distribution of the feedback signal-to-noise ratio; arg is the argument of the independent variable. For example, means that when the function takes the minimum value, the values of variables x and y. In Equation 5, it means the that minimizes takes the value of . refers to the expectation under the corresponding distribution. For example, refers to the expectation under the distribution.

[0029] ; (5) When the image to be processed is transmitted through the image transmission channel, the image to be processed is encoded and decoded using the optimized decoder parameters to complete the encoding and decoding process during the transmission of the image to be processed with the minimum image distortion.

[0030] In this embodiment, the target image is encoded using the feedback signal-to-noise ratio of the image transmission channel to obtain an image coding vector; then, after decoding the image coding vector to obtain a decoded image, based on the target image and the decoded image, the image codec loss is determined. Finally, when the image to be processed is transmitted through the image transmission channel, the image to be processed is encoded and decoded based on the image codec loss to obtain an image transmission result. The present application dynamically adjusts the encoding and decoding process of the image through the channel signal-to-noise ratio, improving the accuracy of image encoding and decoding during image transmission.

[0031] Figure 2 is the second flow diagram of the image transmission method provided by the present invention. As Figure 2 shown, the method may further include: Step 210, determining an input image vector of the target image; the input image vector is determined based on the size of the target image; Step 220, obtaining encoding parameters of a joint channel encoder; the joint channel encoder is used to encode the target image; Step 230, encoding the target image based on the feedback signal-to-noise ratio, the input image vector, and the encoding parameters to obtain an image coding vector.

[0032] Specifically, the size of the target image is represented by height, width, and channels, and the target image is represented by a vector , where and represent the set of integers, and the integers take values from 0 to 255. According to the image compression requirements and the channel signal-to-noise ratio SNR, the algorithms for different convolutional layers and SNR processing layers are determined. The key information of the image data is extracted through the convolutional layer to achieve the effect of image compression and reduce the amount of image data transmission. For example, the size of the CIFAR10 image (a classification dataset composed of color images) data is 3×32×32, and the size of the data to be transmitted after encoding is 64×4×4, which is compressed to one-third of the original image. By using different deep learning network models and parameters, a better compression effect can be achieved. This application reduces the bandwidth required for image data transmission, improves the image transmission speed, and reduces the latency and cost of image data transmission.

[0033] In this embodiment, the channel signal-to-noise ratio is fused with the characteristics of the transmitted image to encode the target image, reducing the image transmission loss.

[0034] In one embodiment, the image transmission method provided by the embodiments of this application may further include: Step 10: Determine the transfer function of the image transmission channel; Step 20: Based on the transfer function, Gaussian distribution samples, and the image coding vector, determine the vector to be decoded; the Gaussian distribution samples are vector samples that conform to the Gaussian distribution; Step 30: Determine the mapping relationship between the vector to be decoded and the feedback signal-to-noise ratio; Step 40: Based on the feedback signal-to-noise ratio, the mapping relationship, and the decoding parameters, decode the vector to be decoded to obtain the decoded image.

[0035] Specifically, the transfer function of the image transmission channel is as shown in the above formula 2, and the vector is determined by independently and identically distributed samples that conform to the Gaussian distribution. Among them, represents the image coding vector; represents the vector to be decoded. The mapping relationship between the vector to be decoded and the feedback signal-to-noise ratio is as shown in the above formula 3, where represents the decoded image.

[0036] In this embodiment, the feedback signal-to-noise ratio participates in the image decoding process and cooperates with the image coding process to improve the accuracy of image encoding and decoding.

[0037] In one embodiment, the image transmission method provided by the embodiments of this application may further include: Step 310: Compare the target image and the decoded image pixel by pixel to obtain the image pixel loss; Step 320: Determine an image discrimination loss based on a first determination probability and a second determination probability; the first determination probability is the probability obtained by the target discriminator for determining the target image; the second determination probability is the probability obtained by the target discriminator for determining the decoded image; Step 330: Determine an image adversarial loss based on a third determination probability and the target image; the third determination probability is the probability obtained by the updated discriminator for determining the decoded image; the updated discriminator is obtained by updating the target discriminator based on the first determination probability and the second determination probability; Step 340: Determine an image smoothing loss based on the decoded image.

[0038] Specifically, the present application also incorporates a Generative Adversarial Networks (GAN), which is a deep learning model composed of a generator network and a discriminator network.

[0039] The main function of the generator network is to create relatively realistic image data samples. The generator network receives a random noise as input, and this noise can be sampled from a Gaussian distribution or other probability distributions. Through the transformation of neural network layers, the generator converts these random noises into data forms, such as images, audio, and text, etc. The goal of the generator is to maximize the probability that the data it generates is misjudged as real data by the discriminator.

[0040] The role of the discriminator network is to distinguish between the fake data generated by the generator and the real data in the real data set. The discriminator network receives the fake data from the generator and the real data from the real data set as input and identifies the differences between the two. The goal of the discriminator is to minimize the probability of misjudging the fake data as real data, while maximizing the correct rate of identifying real data.

[0041] Generator network maps the latent to the data space, while the discriminator network determines the probability that it is an actual training sample is , determines is generated by the model through the probability of generation is . The goal of GAN is to find the best binary classifier to distinguish between real data and generated data by training the generator and the discriminator. In this process, the generator strives to fit the distribution of real data and generate more and more real data to deceive the discriminator. The discriminator constantly learns how to better distinguish between real and fake data in order to see through the data generated by the generator. The goal of this invention is to maximize the binary cross entropy (BCE) of the discriminator and minimize the generator, as shown in Formula 6.

[0042] ; (6) ; (7) p is the model prediction probability, q is the true label 0 or 1, and the loss of the discriminator As shown in Formula 7, is a real loss; is the fake loss. The discriminator is updated based on its performance in distinguishing between real and generated images. The discriminator should output high confidence (1) when it sees a real image and low confidence (0) when it sees a generated image. In this way, the discriminator learns how to distinguish between real and generated images. The discriminator's loss is the sum of the real loss and the fake loss, which helps the discriminator better learn to distinguish between the two types of images and update the discriminator parameters , the update formula is: .

[0043] In order to ensure that the image encoding and decoding results have good visual and quantitative scores, this application proposes a new improved loss function and secondary update mechanism. This application converts the pixel-to-pixel Euclidean loss , smoothing loss , Fighting Losses and the discriminator loss Combined with appropriate update order and weights, a new and more refined loss function and a focused update mechanism are formed.

[0044] In this embodiment, the first determination probability is the probability that the target discriminator determines the target image to be true; the second determination probability is the probability that the target discriminator determines the decoded image to be false. The third determination probability is the probability that the updated discriminator determines the image generated by the generator to be true.

[0045] This embodiment can better adjust the parameters of the codec through the new improved loss function and the secondary update mechanism to achieve low distortion in the image encoding and decoding process.

[0046] In one embodiment, the image transmission method provided in the embodiment of the present application may further include: Step 500: Update the target codec parameters based on at least one of the first update method, the second update method, and the third update method; the first update method is determined based on the image pixel loss; the second update method is determined based on the image adversarial loss and the image smoothing loss; the third update method is determined based on the image pixel loss, the image adversarial loss, and the image smoothing loss.

[0047] Specifically, this application provides three ways to update the codec parameters.

[0048] The first update method: By comparing each pixel of the decoded image with the target image using the Euclidean distance, it is measured whether the noise is correctly filled with pixel colors. This is the loss function of the codec and the most direct evaluation index for measuring the distortion degree of the decoded picture. This index has no direct connection with the discriminator. Therefore, after image decoding, a single loss can be first used to update the encoder and decoder, and the gradient is cleared after the update. As shown in Equation 8. Among them, is the gradient descent update, and the Euclidean loss helps to improve the encoding and decoding methods of the codec at different SNRs, and the adversarial noise gives a decoded image closer to the actual situation at the pixel level.

[0049] ; (8) The second update method: The adversarial loss is to make the decoder generate better outputs to deceive the discriminator. Therefore, the binary cross-entropy of the probability that the discriminator determines that the generated image is true and the true label is used for measurement. The smaller the entropy value, the less able the discriminator is to distinguish whether it is the decoded image or the original image, that is, the better the decoding effect. The adversarial loss is defined as shown in Equation 9, where Dis is the discriminator; Gen is the generator.

[0050] ; (9) A new loss, namely the smoothing loss, is added to these existing loss functions to prevent the main differences between adjacent pixels, which may lead to checkerboard patterns in the image. To calculate the smoothing loss, this application slides a copy of the generated image one pixel unit to the left and down respectively, and then takes the Euclidean distance between the shifted image and the decoded image. The new smoothing loss function is defined as shown in Equation 10, where is the mean square loss of the difference represented by the Euclidean distance in the horizontal direction, is the mean square loss of the difference represented by the Euclidean distance in the vertical direction.

[0051] ; (10) Although the adversarial loss uses the discriminator, it measures the decoded image and should be used to improve the decoder. Also, since the encoder and decoder are an integrated whole, simply updating the decoder parameters may lead to incompatibility with the encoder and affect the image decoding effect. Therefore, the adversarial loss will be used to update both the encoder and decoder simultaneously. The same applies to the smoothing loss. The second update method of this application updates the parameters of the encoder and decoder based on the weighted sum of the adversarial loss and the smoothing loss, as shown in Equation 11.

[0052] ; (11) In this way, the training emphasis can be adjusted more flexibly according to actual needs. If smoother and more natural images are to be generated, the loss coefficient of the smoothing loss is increased. If the patterns of the images themselves are complex, the loss coefficient of the adversarial loss is increased.

[0053] The third update method: Define a new total loss function and update the encoder and decoder only once per iteration, as shown in Equation 12. At initialization, it can be determined .

[0054] ; (12) In this embodiment, the parameters of the encoder and decoder are updated through multiple update methods, which can better adjust the parameters of the encoder and decoder to achieve low distortion in the image encoding and decoding process.

[0055] In one embodiment, the image transmission method provided by the embodiments of this application may further include: Step 410: When transmitting the image to be processed through the image transmission channel, perform encoding and decoding processing on the image to be processed based on the updated encoding and decoding parameters to obtain an image transmission result; the updated encoding and decoding parameters are obtained by updating the target encoding and decoding parameters based on the update method corresponding to the feedback signal-to-noise ratio.

[0056] Specifically, when evaluating the encoding and decoding parameters, the difference between the decoded image and the original image can be measured from two aspects by the Peak Signal-to-Noise Ratio (PSNR) and the Learned Perceptual Image Patch Similarity (LPIPS). The mean squared error (MSE) of N transmitted images is defined as shown in Equation 13, where and represent the i-th image and its corresponding decoded image, respectively; is the distance defined in MSE, and N is the number of image samples.

[0057] ; (13) In practical applications, assuming a given signal-to-noise ratio (SNR) distribution and uniformly transmitting images in the dataset, the calculation of PSNR is shown in Equation (14), where MAX is the maximum possible value of the image pixels (for example, MAX for an 8-bit color image is 255). First, calculate the PSNR of each image, and then average all the test images. PSNR is a pixel-level error metric that quantifies the difference between the original image and the reconstructed image by calculating the mean squared error (MSE) between them, usually expressed in decibels (dB). A higher value indicates better image quality and a smaller pixel difference between the two images. However, PSNR does not always accurately reflect the human eye's perception of image quality because it does not consider the sensitivity differences of the human eye to different image features.

[0058] ; (14) At different SNRs, update the parameters of the codec using the above different (SNR-corresponding) update methods, and perform encoding and decoding operations on the image to be processed through the updated codec to complete the encoding and decoding transmission of the image to be processed.

[0059] In this embodiment, the codec parameters are updated in different ways to improve the accuracy of the image encoding and decoding process.

[0060] Next, the image transmission device provided by the present invention will be described. The image transmission device described below can be correspondingly referred to the image transmission method described above.

[0061] Please refer to Figure 3 , the present invention also provides an image transmission device, including: A feedback SNR acquisition module 301 for acquiring the feedback SNR of the image transmission channel; An image coding vector determination module 302 for encoding a target image based on the feedback SNR to obtain an image coding vector; An image codec loss determination module 303 for determining the image codec loss based on the target image and the decoded image; the decoded image is obtained by decoding the image coding vector; An image transmission module 304 for, when transmitting an image to be processed through the image transmission channel, performing encoding and decoding operations on the image to be processed based on the image codec loss to obtain an image transmission result.

[0062] Optionally, the image coding vector determination module includes: An input image vector determination unit for determining the input image vector of the target image; the input image vector is determined based on the size of the target image; A coding parameter acquisition unit for acquiring the coding parameters of the joint channel encoder; the joint channel encoder is used for encoding the target image; An image coding vector determination unit for encoding the target image based on the feedback signal-to-noise ratio, the input image vector, and the coding parameters to obtain an image coding vector.

[0063] Optionally, the image transmission device further includes: A transmission function determination module for determining the transmission function of the image transmission channel; A vector to be decoded determination module for determining a vector to be decoded based on the transmission function, a Gaussian distribution sample, and the image coding vector; the Gaussian distribution sample is a vector sample conforming to a Gaussian distribution; A mapping relationship determination module for determining the mapping relationship between the vector to be decoded and the feedback signal-to-noise ratio; A vector to be decoded decoding module for decoding the vector to be decoded based on the feedback signal-to-noise ratio, the mapping relationship, and decoding parameters to obtain the decoded image.

[0064] Optionally, determining the image coding and decoding loss based on the target image and the decoded image includes: A pixel comparison unit for performing pixel comparison between the target image and the decoded image to obtain an image pixel loss; An image discrimination loss determination unit for determining an image discrimination loss based on a first determination probability and a second determination probability; the first determination probability is the probability obtained by the target discriminator for determining the target image; the second determination probability is the probability obtained by the target discriminator for determining the decoded image; An image adversarial loss determination unit for determining an image adversarial loss based on a third determination probability and the target image; the third determination probability is the probability obtained by the updated discriminator for determining the decoded image; the updated discriminator is obtained by updating the target discriminator based on the first determination probability and the second determination probability; An image smoothing loss determination unit for determining an image smoothing loss based on the decoded image.

[0065] Optionally, the image transmission device further includes: A target encoding / decoding parameter update module is used to update target encoding / decoding parameters based on at least one of a first update method, a second update method, and a third update method; the first update method is determined based on the image pixel loss; the second update method is determined based on the image adversarial loss and the image smoothing loss; the third update method is determined based on the image pixel loss, the image adversarial loss, and the image smoothing loss.

[0066] Optionally, the image transmission module includes: An image transmission unit, configured to perform encoding / decoding processing on a to-be-processed image based on updated encoding / decoding parameters to obtain an image transmission result when transmitting the to-be-processed image through the image transmission channel; the updated encoding / decoding parameters are obtained by updating the target encoding / decoding parameters based on an update method corresponding to the feedback signal-to-noise ratio.

[0067] Figure 4 An example of a schematic physical structure diagram of an electronic device is shown as Figure 4 shown. The electronic device may include: a processor 410, a communication interface 420, a memory 430, and a communication bus 440. Among them, the processor 410, the communication interface 420, and the memory 430 communicate with each other through the communication bus 440. The processor 410 can call logic instructions in the memory 430 to execute an image transmission method, which includes: obtaining the feedback signal-to-noise ratio of an image transmission channel; encoding a target image based on the feedback signal-to-noise ratio to obtain an image encoding vector; determining an image encoding / decoding loss based on the target image and a decoded image; the decoded image is obtained by decoding the image encoding vector; when transmitting a to-be-processed image through the image transmission channel, performing encoding / decoding processing on the to-be-processed image based on the image encoding / decoding loss to obtain an image transmission result.

[0068] In addition, when the logical instructions in the above-mentioned memory 430 are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs that can store program codes.

[0069] On the other hand, the present invention also provides a computer program product. The computer program product includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the image transmission method provided by the above-mentioned various methods. The method includes: obtaining the feedback signal-to-noise ratio of an image transmission channel; encoding a target image based on the feedback signal-to-noise ratio to obtain an image coding vector; determining an image coding and decoding loss based on the target image and a decoded image, where the decoded image is obtained by decoding the image coding vector; and when transmitting a to-be-processed image through the image transmission channel, performing coding and decoding processing on the to-be-processed image based on the image coding and decoding loss to obtain an image transmission result.

[0070] On another aspect, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it is configured to execute the image transmission method provided by the above-mentioned various methods. The method includes: obtaining the feedback signal-to-noise ratio of an image transmission channel; encoding a target image based on the feedback signal-to-noise ratio to obtain an image coding vector; determining an image coding and decoding loss based on the target image and a decoded image, where the decoded image is obtained by decoding the image coding vector; and when transmitting a to-be-processed image through the image transmission channel, performing coding and decoding processing on the to-be-processed image based on the image coding and decoding loss to obtain an image transmission result.

[0071] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0072] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0073] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An image transmission method, characterized in that: include: Obtaining feedback signal-to-noise ratio of the image transmission channel; Encoding the target image based on the feedback signal-to-noise ratio to obtain an image coding vector; Determine the image encoding and decoding loss based on the target image and the decoded image; the decoded image is obtained by decoding the image encoding vector; In the case of transmitting the image to be processed through the image transmission channel, encoding and decoding processing is performed on the image to be processed based on the image encoding and decoding loss to obtain an image transmission result.

2. The image transmission method according to claim 1, characterized in that: The encoding of the target image based on the feedback signal-to-noise ratio to obtain an image coding vector comprises: Determine an input image vector of a target image; the input image vector is determined based on a size of the target image; Acquiring encoding parameters of a joint channel encoder; the joint channel encoder is used to encode the target image; Based on the feedback signal-to-noise ratio, the input image vector and the encoding parameter, the target image is encoded to obtain an image encoding vector.

3. The image transmission method according to claim 1, characterized in that: The image transmission method further includes: determining a transfer function of the image transmission channel; Determine a vector to be decoded based on the transfer function, Gaussian distribution samples and the image coding vector; the Gaussian distribution samples are vector samples that conform to Gaussian distribution; Determining a mapping relationship between the to-be-decoded vector and the feedback signal-to-noise ratio; Based on the feedback signal-to-noise ratio, the mapping relationship and the decoding parameters, the vector to be decoded is decoded to obtain the decoded image.

4. The image transmission method according to claim 1, characterized in that: The determining of the image encoding and decoding loss based on the target image and the decoded image comprises: Comparing the target image and the decoded image pixel by pixel to obtain image pixel loss; Determine the image discrimination loss based on a first judgment probability and a second judgment probability; the first judgment probability is the probability obtained by the target discriminator judging the target image; the second judgment probability is the probability obtained by the target discriminator judging the decoded image; Determine the image adversarial loss based on a third determination probability and the target image; the third determination probability is a probability obtained by updating the discriminator to determine the decoded image; the updated discriminator is obtained by updating the target discriminator based on the first determination probability and the second determination probability; An image smoothing loss is determined based on the decoded image.

5. The image transmission method according to claim 4, characterized in that: The step of determining the image encoding and decoding loss based on the target image and the decoded image comprises: The target encoding and decoding parameters are updated based on at least one of a first updating method, a second updating method and a third updating method; the first updating method is determined based on the image pixel loss; the second updating method is determined based on the image adversarial loss and the image smoothness loss; the third updating method is determined based on the image pixel loss, the image adversarial loss and the image smoothness loss.

6. The image transmission method according to claim 5, characterized in that: In the case of transmitting the image to be processed through the image transmission channel, encoding and decoding the image to be processed based on the image encoding and decoding loss to obtain the image transmission result includes: When transmitting the image to be processed through the image transmission channel, encoding and decoding processing is performed on the image to be processed based on updated encoding and decoding parameters to obtain an image transmission result; the updated encoding and decoding parameters are obtained by updating the target encoding and decoding parameters based on an update method corresponding to the feedback signal-to-noise ratio.

7. An image transmission device, characterized in that: include: A feedback signal-to-noise ratio acquisition module, used to acquire the feedback signal-to-noise ratio of the image transmission channel; An image coding vector determination module, used to encode the target image based on the feedback signal-to-noise ratio to obtain an image coding vector; An image coding loss determination module, used to determine the image coding loss based on the target image and a decoded image; the decoded image is obtained by decoding the image coding vector; The image transmission module is used to perform encoding and decoding processing on the image to be processed based on the image encoding and decoding loss when transmitting the image to be processed through the image transmission channel to obtain an image transmission result.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the image transmission method according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the image transmission method according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the image transmission method according to any one of claims 1 to 6 is implemented.

Images