Image loss determination method and related apparatus

Abstract

The embodiments of the present application provide an image loss determination method and a related apparatus. The method comprises: generating a standard dynamic range (SDR) image on the basis of an original high dynamic range (HDR) image; generating a reconstructed HDR image on the basis of the original HDR image and the SDR image; and determining image losses of a plurality of regions in the reconstructed HDR image on the basis of the original HDR image and the reconstructed HDR image. Using the embodiments of the present application can obtain reconstruction losses of HDR images, thereby identifying visual contours in reconstructed HDR images.

Description

Image loss determination method and related device

[0001] The present application claims priority to the Chinese patent application No. 202410745724.2, filed on June 7, 2024, and entitled "Image loss determination method and related device", the whole content of which is incorporated herein by reference. TECHNICAL FIELD

[0002] The present application relates to the field of image processing, in particular to an image loss determination method and related device. BACKGROUND

[0003] In the field of digital image processing, HDR (High Dynamic Range) images have attracted much attention due to their ability to capture and display a wide range of luminance in the real world. However, due to technical and equipment limitations, we often can only obtain SDR (Standard Dynamic Range) images, which have limited luminance range and cannot fully display the rich details of the real world. To overcome this limitation, the technical personnel in the field proposed a method of reconstructing HDR images from SDR images and gain maps. The gain map is a special image that records the luminance gain information required for each pixel point from the SDR image to the HDR image. The gain map can be regarded as a two-dimensional matrix, where each element corresponds to the luminance gain value of a pixel point in the SDR image.

[0004] In the process of reconstructing the HDR image, the SDR image needs to be preprocessed, including denoising, contrast enhancement and other operations, to improve the quality of the image. Then, according to the gain map, the luminance of each pixel point in the SDR image is adjusted. Specifically, for each pixel point in the SDR image, we find the gain value at the corresponding position in the gain map and apply the gain value to the pixel luminance value of the SDR image. In this way, we get an SDR image whose luminance range has been expanded towards the direction of the HDR image. In the process of reconstructing the HDR image, further optimization of the color and details of the image is usually required. The fusion of the SDR image and the SDR image adjusted by the gain map can obtain a result closer to the HDR image.

[0005] Reconstructing the HDR image from the SDR image and the gain map is an effective digital image processing technology, but if the gain map has a large gain range or a small bit depth, there will be loss in the restoration process, resulting in visible contours in the final restored image. How to identify the visible contours is a technical problem that the technical personnel in the field are studying. SUMMARY

[0006] The embodiment of the present application discloses an image loss determination method and related device, which can obtain the reconstruction loss of the HDR image, thereby identifying the visible fault in the reconstructed HDR image.

[0007] In the first aspect, the embodiment of the present application provides an image loss determination method, which comprises:

[0008] generating a standard dynamic range (SDR) image according to an original high dynamic range (HDR) image;

[0009] generating a reconstructed HDR image according to the original HDR image and the SDR image;

[0010] determining image losses of a plurality of regions in the reconstructed HDR image according to the original HDR image and the reconstructed HDR image.

[0011] By using the method, the SDR image and the reconstructed HDR image are obtained in sequence through a series of processing on the original HDR image, and then the image loss is generated according to the original HDR image and the reconstructed HDR image. The image loss can be used as feedback to optimize and adjust the image processing process, or the image loss can be used to evaluate the reconstruction process, such as identifying the visible fault in the reconstructed HDR image according to the image loss. Therefore, the image loss is beneficial to determining the image state or improving the image processing effect.

[0012] With reference to the first aspect, in a possible implementation manner of the first aspect, the plurality of regions comprises a plurality of pixels.

[0013] With reference to the first aspect, or any of the possible implementation manners of the first aspect, in another possible implementation manner of the first aspect, the method further comprises:

[0014] determining regions meeting a loss threshold and regions not meeting the loss threshold in the plurality of regions according to the image losses of the plurality of regions, wherein the loss threshold is a minimum value of an image brightness that can be perceived by naked eyes.

[0015] That is, the image loss can be used to determine which regions in the plurality of regions have acceptable reconstruction loss and which regions have unacceptable reconstruction loss, and the image loss can reflect the overall reconstruction effect of the image, which is beneficial to the user to properly optimize the subsequent image processing process.

[0016] With reference to the first aspect, or any of the possible implementation manners of the first aspect, in another possible implementation manner of the first aspect, the method further comprises:

[0017] generating the loss threshold according to information in the original HDR image.

[0018] That is to say, the loss threshold for measuring whether the image loss is qualified is determined according to the original HDR image, that is, a loss threshold is generated for each HDR image, instead of a fixed loss threshold. The threshold calculated in this way is more targeted when measuring the reconstruction effect of the reconstructed HDR image, and therefore the judgment result is more accurate.

[0019] With reference to the first aspect or any possible implementation manner of the first aspect, in a possible implementation manner of the first aspect, the generating the loss threshold according to information in the original HDR image comprises:

[0020] inputting parameter information in the original HDR image into a Barten model to generate the loss threshold, wherein the parameter information comprises one or more of neural noise, lateral inhibition, photon noise, external noise, limited integration capacity, optical modulation transfer function, direction and temporal filtering.

[0021] It can be understood that, by using the Barten model, many factors of human eye contrast perception are considered, which is equivalent to adjusting according to the actual scene, and the loss threshold obtained has higher accuracy.

[0022] With reference to the first aspect or any possible implementation manner of the first aspect, in a possible implementation manner of the first aspect, the inputting parameter information in the original HDR image into a Barten model to generate the loss threshold comprises:

[0023] inputting personalized information and parameter information in the original HDR image into a Barten model to generate the loss threshold, wherein the personalized information comprises one or more of age, gender, type, vision of a user group of the original HDR image.

[0024] It can be understood that, in this implementation manner, the generation of the loss threshold not only considers the characteristics of the original HDR image, but also considers the perception characteristics of the user group of the HDR image to image information, and therefore the loss threshold generated has higher accuracy and better effect when measuring whether the image loss can be accepted by the user group.

[0025] With reference to the first aspect or any possible implementation manner of the first aspect, in a possible implementation manner of the first aspect, the generating the loss threshold according to information in the original HDR image comprises:

[0026] generating the loss threshold using square root uniformity in dark places of the original HDR image and using logarithmic uniformity in bright places of the original HDR image by using a Schreiber model.

[0027] It can be understood that, in this scheme, the contrast perception of the human eye can be described by a model according to different visual characteristics of the human eye in the dark and the light, and the calculation has a faster speed and better practicability.

[0028] With reference to the first aspect or any possible implementation of the first aspect, in a possible implementation of the first aspect, the method further includes:

[0029] generating an image loss map, wherein the loss map includes a plurality of regions having the same distribution rule as the original HDR image, and a luminance value of a region satisfying a loss threshold in the plurality of regions of the loss map is marked as a first value, and a luminance value of a region not satisfying the loss threshold is marked as a second value.

[0030] It can be understood that, through the loss map, the user or the developer can directly view the regions satisfying the image loss condition and the regions not satisfying the condition, which is beneficial to the user to quickly and accurately perform subsequent related operations.

[0031] With reference to the first aspect or any possible implementation of the first aspect, in a possible implementation of the first aspect, the method further includes:

[0032] performing luminance conversion processing on the original HDR image to obtain a first image;

[0033] performing luminance conversion processing on the reconstructed HDR image to obtain a second image;

[0034] determining the image luminance loss of the plurality of regions in the HDR image according to the first image and the second image.

[0035] With reference to the first aspect or any possible implementation of the first aspect, in a possible implementation of the first aspect, the method further includes:

[0036] performing pixel value transformation processing on the original HDR image to obtain a reference image;

[0037] performing luminance conversion processing on the reference image to obtain a first image;

[0038] performing luminance conversion processing on the reconstructed HDR image to obtain a second image;

[0039] determine image brightness loss of a plurality of regions in the HDR image according to the first image and the second image.

[0040] In this way, the original HDR image is first subjected to pixel value transformation processing to obtain a reference image; subsequently, the image loss between the reconstructed HDR image and the reference image is determined, and the image loss is better for judging whether the transition appears discontinuity when the brightness changes.

[0041] With reference to the first aspect, or any possible implementation manner of the first aspect, in a possible implementation manner of the first aspect, the pixel value transformation processing on the original HDR image comprises:

[0042] one or more of the following operations are performed:

[0043] the R value of each pixel point in the original HDR image is increased by a first offset to obtain a reference image;

[0044] the G value of each pixel point in the original HDR image is increased by a second offset to obtain a reference image;

[0045] the B value of each pixel point in the original HDR image is increased by a third offset to obtain a reference image;

[0046] the R value of each pixel point in the original HDR image is subtracted by a first offset to obtain a reference image;

[0047] the G value of each pixel point in the original HDR image is subtracted by a second offset to obtain a reference image;

[0048] the B value of each pixel point in the original HDR image is subtracted by a third offset to obtain a reference image.

[0049] With reference to the first aspect, or any possible implementation manner of the first aspect, in a possible implementation manner of the first aspect, the method further comprises:

[0050] if the regions that do not meet the loss threshold in the plurality of regions exceed a preset proportion, adjusting model parameters, wherein the model parameters comprise parameters required for generating a reconstructed HDR image according to the HDR image and the SDR image;

[0051] returning to perform the step of generating a reconstructed HDR image according to the HDR image and the SDR image according to the model parameters.

[0052] It can be understood that taking the image loss as the basis for adjusting the model parameters is beneficial to subsequently reconstructing a better HDR image.

[0053] With reference to the first aspect, or any possible implementation of the first aspect, in a possible implementation of the first aspect, the generating a reconstructed HDR image according to the original HDR image and the SDR image comprises:

[0054] generating a gain map according to the original HDR image and the SDR image;

[0055] generating the reconstructed HDR image according to the gain map and the SDR image.

[0056] It can be seen that the gain map is generated in the process of reconstructing the HDR image, and the gain map can be used to enhance the dynamic range of the SDR image, so as to obtain the effect of the HDR image.

[0057] In a second aspect, an embodiment of the present application provides an image loss determination apparatus, which comprises:

[0058] a first generating unit configured to generate a standard dynamic range (SDR) image according to an original high dynamic range (HDR) image;

[0059] a second generating unit configured to generate a reconstructed HDR image according to the original HDR image and the SDR image;

[0060] a first determining unit configured to determine image loss of a plurality of regions in the reconstructed HDR image according to the original HDR image and the reconstructed HDR image.

[0061] By using this method, the SDR image and the reconstructed HDR image are obtained in sequence through a series of processing on the original HDR image, and then the image loss is generated according to the original HDR image and the reconstructed HDR image. The image loss can be used as feedback to optimize and adjust the image processing process, or the image loss can be used to evaluate the reconstruction process, such as identifying the visible fault in the reconstructed HDR image according to the image loss. Therefore, with the image loss, it is beneficial to determine the image state or improve the image processing effect.

[0062] With reference to the second aspect, in a possible implementation of the second aspect, the plurality of regions comprises a plurality of pixels.

[0063] With reference to the second aspect, or any possible implementation of the second aspect, in a possible implementation of the second aspect, the apparatus further comprises:

[0064] The second determining unit is configured to determine, according to the image loss of the plurality of regions, a region satisfying a loss threshold and a region not satisfying the loss threshold in the plurality of regions, wherein the loss threshold is a minimum value of an estimated image brightness that can be perceived by the naked eye.

[0065] That is, it can be determined by the image loss which regions in the plurality of regions have acceptable reconstruction loss and which regions do not have acceptable reconstruction loss, and the reconstruction effect of the entire image can be reflected as a whole, which is beneficial to the user to properly optimize the subsequent image processing process.

[0066] With reference to the second aspect or any possible implementation manner of the second aspect, in a possible implementation manner of the second aspect, the apparatus further includes:

[0067] The third generating unit is configured to generate the loss threshold according to information in the original HDR image.

[0068] That is, the loss threshold for measuring the image loss is determined according to the original HDR image, that is, a loss threshold is generated for each HDR image, rather than a fixed loss threshold. The loss threshold thus calculated is more targeted when measuring the reconstruction effect of the reconstructed HDR image, and therefore the judgment result is more accurate.

[0069] With reference to the second aspect or any possible implementation manner of the second aspect, in a possible implementation manner of the second aspect, in the aspect of generating the loss threshold according to information in the original HDR image, the third generating unit is specifically configured to:

[0070] input parameter information in the original HDR image into a Barten model to generate the loss threshold, wherein the parameter information includes one or more of neural noise, lateral inhibition, photon noise, external noise, limited integration capacity, optical modulation transfer function, direction, and temporal filtering.

[0071] It can be understood that, by using the Barten model, many factors of the contrast perception of the human eye are considered, which is equivalent to adjusting according to the actual scene, and the loss threshold thus obtained has higher accuracy.

[0072] With reference to the second aspect or any possible implementation manner of the second aspect, in a possible implementation manner of the second aspect, in the aspect of inputting the parameter information in the original HDR image into the Barten model to generate the loss threshold, the third generating unit is specifically configured to:

[0073] inputting personalized information and parameter information in the original HDR image into the Barton model to generate the loss threshold, wherein the personalized information comprises one or more of age, gender, type, and vision of a user group of the original HDR image.

[0074] It can be understood that in this implementation, the generation of the loss threshold not only considers the characteristics of the original HDR image, but also considers the perception characteristics of the user group of the HDR image to the image information, so that the generated loss threshold has higher result accuracy and better effect when measuring whether the image loss can be accepted by the user group.

[0075] With reference to the second aspect or any possible implementation manner of the second aspect, in a possible implementation manner of the second aspect, the third generation unit is configured to:

[0076] generating the loss threshold using square root uniformity in dark places of the original HDR image and using logarithmic uniformity in bright places of the original HDR image through the Schreiber model.

[0077] It can be understood that in this scheme, the different visual characteristics of the human eye in dark places and bright places can be used to basically describe the contrast perception of the human eye by using the model, which has faster calculation speed and better practicability.

[0078] With reference to the second aspect or any possible implementation manner of the second aspect, in a possible implementation manner of the second aspect, the apparatus further comprises:

[0079] a fourth generation unit configured to generate an image loss schematic diagram, wherein the loss schematic diagram comprises a plurality of regions having the same distribution rule as the original HDR image, and in the plurality of regions of the loss schematic diagram, a luminance value of a region satisfying the loss threshold is marked as a first value, and a luminance value of a region not satisfying the loss threshold is marked as a second value.

[0080] It can be understood that the user or the developer can directly view the regions satisfying the image loss condition and the regions not satisfying the image loss condition through the loss schematic diagram, which is beneficial to the user to quickly and accurately perform subsequent related operations.

[0081] With reference to the second aspect or any possible implementation manner of the second aspect, in a possible implementation manner of the second aspect, the first determination unit is configured to:

[0082] performing brightness conversion processing on the original HDR image to obtain a first image;

[0083] performing brightness conversion processing on the reconstructed HDR image to obtain a second image;

[0084] determining image brightness loss of a plurality of regions in the HDR image according to the first image and the second image.

[0085] With reference to the second aspect, or any possible implementation manner of the second aspect, in a possible implementation manner of the second aspect, in the aspect of determining image loss of a plurality of regions in the HDR image according to the original HDR image and the reconstructed HDR image, the first determining unit is specifically configured to:

[0086] performing pixel value transformation processing on the original HDR image to obtain a reference image;

[0087] performing brightness conversion processing on the reference image to obtain a first image;

[0088] performing brightness conversion processing on the reconstructed HDR image to obtain a second image;

[0089] determining image brightness loss of a plurality of regions in the HDR image according to the first image and the second image.

[0090] In this manner, the original HDR image is first subjected to pixel value transformation processing to obtain a reference image, and then image loss between the reconstructed HDR image and the reference image is determined, which is better for determining whether a discontinuity occurs in the transition when brightness changes.

[0091] With reference to the second aspect, or any possible implementation manner of the second aspect, in a possible implementation manner of the second aspect, in the aspect of performing pixel value transformation processing on the original HDR image to obtain a reference image, the first determining unit is specifically configured to:

[0092] performing one or more of the following operations:

[0093] increasing the R value of each pixel point in the original HDR image by a first offset to obtain a reference image;

[0094] increasing the G value of each pixel point in the original HDR image by a second offset to obtain a reference image;

[0095] increasing the B value of each pixel point in the original HDR image by a third offset to obtain a reference image;

[0096] Subtracting a first offset from the R value of each pixel point in the original HDR image to obtain a reference image;

[0097] Subtracting a second offset from the G value of each pixel point in the original HDR image to obtain a reference image;

[0098] Subtracting a third offset from the B value of each pixel point in the original HDR image to obtain a reference image.

[0099] With reference to the second aspect, or any possible implementation of the second aspect, in a possible implementation of the second aspect, the apparatus further includes:

[0100] The adjusting unit is configured to adjust a model parameter in a case where a region that does not satisfy the loss threshold in the plurality of regions exceeds a preset proportion, wherein the model parameter includes a parameter required for generating a reconstructed HDR image according to the HDR image and the SDR image.

[0101] The second generating unit is triggered to return to execute the step of generating the reconstructed HDR image according to the HDR image and the SDR image according to the model parameter.

[0102] It can be understood that taking the image loss as a basis for adjusting the model parameter is beneficial to subsequently reconstructing a better HDR image.

[0103] With reference to the second aspect, or any possible implementation of the second aspect, in a possible implementation of the second aspect, the second generating unit is specifically configured to:

[0104] Generate a gain map according to the original HDR image and the SDR image.

[0105] Generate a reconstructed HDR image according to the gain map and the SDR image.

[0106] It can be seen that the gain map is generated first in the process of reconstructing the HDR image, and the gain map can be used to enhance the dynamic range of the SDR image, thereby obtaining the effect of the HDR image.

[0107] In a third aspect, an electronic device is provided, which includes one or more processors, a memory, wherein the memory is coupled to the one or more processors, the memory is configured to store a computer program, and the one or more processors are configured to invoke the computer program to enable the electronic device to perform the method described in the first aspect or any possible implementation of the first aspect.

[0108] In a fourth aspect, an embodiment of the present application provides a chip system applied to an electronic device, the chip system comprising one or more processors, the processor being configured to invoke a computer program to enable the electronic device to perform the method described in the first aspect or any possible implementation of the first aspect.

[0109] In a fifth aspect, an embodiment of the present application provides a computer program product comprising instructions, which, when executed on an electronic device, enable the electronic device to perform the method described in the first aspect or any possible implementation of the first aspect.

[0110] In a sixth aspect, an embodiment of the present application provides a computer-readable storage medium, including a computer program, which, when executed on an electronic device, enables the electronic device to perform the method described in the first aspect or any possible implementation of the first aspect. BRIEF DESCRIPTION OF DRAWINGS

[0111] The following describes the drawings used in the embodiments of the present application.

[0112] FIG. 1 is a structural schematic diagram of an electronic device according to an embodiment of the present application;

[0113] FIG. 2 is a flowchart of an image loss determination method according to an embodiment of the present application;

[0114] FIG. 3 is a flowchart of a reconstructed HDR image according to an embodiment of the present application;

[0115] FIG. 4 is an effect diagram of a reconstructed HDR image according to an embodiment of the present application;

[0116] FIG. 5 is a flowchart of an image loss determination method according to an embodiment of the present application;

[0117] FIG. 6 is a generation diagram of a reference image according to an embodiment of the present application;

[0118] FIG. 7 is a generation diagram of a reference image according to an embodiment of the present application;

[0119] FIG. 8 is a diagram of not generating a reference image according to an embodiment of the present application;

[0120] FIG. 9 is a flowchart of determining whether the image loss meets the loss threshold according to an embodiment of the present application;

[0121] FIG. 10 is a diagram of the relationship between the loss threshold and the image brightness according to an embodiment of the present application;

[0122] FIG. 11 is a loss diagram according to an embodiment of the present application;

[0123] FIG. 12 is a schematic diagram of an image loss determination device according to an embodiment of the present application. DETAILED DESCRIPTION

[0124] The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0125] Referring to FIG. 1, FIG. 1 is a schematic diagram of an electronic device according to an embodiment of the present application. The electronic device 10 can be a single device or a device cluster composed of multiple devices. The electronic device 10 has a certain computing capability and can process images. The electronic device 10 can be a cloud device (such as a cloud server) or a local device.

[0126] For example, the electronic device 10 can be a server, a computer, a handheld device (for example, a mobile phone, a tablet computer, a palm computer, etc.), a vehicle-mounted device (for example, a car, a bicycle, an electric vehicle, an airplane, a ship, etc.), a wearable device (for example, a smart watch (such as iWatch, etc.), a smart bracelet, a pedometer, etc.), a smart home device (for example, a refrigerator, a television, an air conditioner, a power meter, etc.), a smart robot, a workshop device, etc., as long as it has image processing or reality requirements.

[0127] The electronic device 10 includes one or more processors 101 and one or more memories 102. The processor 101 and the memory 102 can be connected by a bus or other means, wherein:

[0128] The memory 102 stores a computer program. The processor 101 is configured to read the computer program in the memory 102 and execute the method defined by the computer program, thereby realizing the image processing function, such as generating an SDR image according to an original HDR image, generating a gain map according to the original HDR image and the SDR image, generating a reconstructed HDR image according to the SDR image and the gain map; optionally, the loss threshold can be further calculated by calling the computer program, and whether the image loss meets the condition can be evaluated by the loss threshold, etc.

[0129] More specifically, the processor 101 can call the computer program in the memory 102 to execute one or more steps in steps S201-S208 shown in FIG. 2.

[0130] The processor 101 can include a graphics processing unit (GPU), can also include a central processing unit (CPU), and can further include other types of processors. Here, the processor 101 can be of the same type or of different types, and can be integrated on the same chip or can be independent chips.

[0131] In addition, the memory 102 also stores other data in addition to the computer program. The other data can include data required during the running of the processor, such as the original HDR image, and can also include data generated during the running of the processor, such as the SDR image, the gain map, the reconstructed HDR image, the image loss, the metadata of the related image, and the like. Of course, the other data can also include other data, which is not listed here.

[0132] The memory 102 generally includes an internal memory and an external memory. The internal memory can be a random access memory (RAM), a read-only memory (ROM), and a cache (CACHE), etc. The external memory can be a hard disk, an optical disk, a USB disk, a floppy disk, or a tape drive, etc. The computer program is generally stored on the external memory, and the processor loads the computer program from the external memory to the internal memory before executing the processing. The input image (such as the original HDR image) required by the processor in the embodiment of the present application can be stored on the external memory, and the image can be loaded to the internal memory when needed.

[0133] In the embodiment of the present application, the electronic device 10 can also include other devices or modules, such as an audio input / output module, an image acquisition module (such as a camera), an image output device (such as a display screen), a communication module (for receiving or sending related data), an instruction input / output module (such as a keyboard, a touch screen), and the like. The specific devices or modules included can be matched according to the specific use scenario.

[0134] Please refer to FIG. 2, which is a flowchart of an image loss determination method provided by an embodiment of the present application. The method can be implemented based on the electronic device 10 shown in FIG. 1, or can be implemented based on other architectures. The method includes but is not limited to the following steps:

[0135] Step S201: generating a standard dynamic range (SDR) image according to an original high dynamic range (HDR) image.

[0136] Specifically, there are many ways to generate SDR images from HDR images, for example, the following process can be used: convert the HDR nonlinear electrical signal to the HDR linear optical signal through the electro-optical transfer function (EOTF), then do color space conversion (Color Space Converting) on the HDR linear optical signal, such as converting from BT.2020 to BT.709, then tone map (Tone Mapping) the HDR linear optical signal to the SDR linear optical signal, and then convert the SDR linear optical signal to the SDR nonlinear electrical signal through the optical-electro transfer function (OETF). In this implementation process, the conversion of the HDR image to the SDR image undergoes compression of the luminance dynamic range and the color space, for example, the luminance range is compressed from [0.0005, 10000] nit to [0.1, 100] nit, and the color space is converted from BT.2020 to BT.709; at the same time, the color depth is reduced from 10 bits to 8 bits; the number of available color steps of the video signal is reduced from 1024 to 256, a reduction of 75%; at the same time, the electro-optical transfer function EOTF is changed from perceptual quantization (PQ) or hybrid log-gamma (HLG) to the BT.709 gamma function.

[0137] In the embodiments of the present application, the HDR image can also be referred to as a pre-enhanced image, and the SDR image can also be referred to as a base image. It should be noted that there are many ways to convert the HDR image to the SDR image, which are not listed here.

[0138] Step S202: generating a reconstructed HDR image according to the original HDR image and the SDR image.

[0139] Specifically, the HDR image can be reconstructed based on the SDR image, and some information in the original HDR image can be used as reference information or constraint in the reconstruction process. Optionally, as shown in FIG. 3, the reconstruction process can include the following operations: generating a gain map Gainmap based on the original HDR image and the SDR image, optionally, generating a floating-point gain map Gainmap(float) based on the original HDR image and the SDR image, converting the floating-point gain map Gainmap(float) into an integer gain map NormalizeGainmap(int), i.e., float 32=>int 8, and then converting the integer gain map NormalizeGainmap(int) into a floating-point gain map Gainmap(float), i.e., int 8=>float 32, to obtain a final gain map, and then generating the reconstructed HDR image based on the gain map and the SDR image. Optionally, in this process, the gain map can be used to enhance the dynamic range of the SDR image, so as to obtain the effect of the HDR image.

[0140] Optionally, the following formula is used to generate the gain map Gainmap:

[0141] wherein, Alternate is the HDR image linear space luminance, Baseline is the SDR image linear space luminance, K alternate is the HDR image linear space offset, K baseline is the SDR image linear space offset.

[0142] Step S203: determining image loss of multiple regions in the reconstructed HDR image based on the original HDR image and the reconstructed HDR image.

[0143] As shown in FIG. 3, the reconstructed HDR image is compared with the original HDR image to determine the image loss of multiple regions in the reconstructed HDR image. Optionally, the multiple regions include multiple pixels, i.e., the regions are divided in the granularity of pixel points; optionally, the multiple regions include multiple pixel point sets, i.e., the regions are divided in the granularity of pixel point sets, each pixel point set includes multiple pixel points, and the number of pixel points included in the pixel point set can be pre-configured according to needs and application scenarios, which is not limited herein.

[0144] Optionally, the image loss can include loss of one or more of luminance, chrominance, contrast, etc.

[0145] As shown in FIG. 4, part (a) shows a reconstructed HDR image, and part (b) is a local enlarged view of the reconstructed HDR image. As can be seen from part (b), there are black and white wavy lines near the sun, which is caused by image loss (e.g. luminance loss) in the process of reconstructing the HDR image. The image loss can be calculated.

[0146] For ease of understanding, several ways of determining the image loss are listed below.

[0147] Case 1: The method for determining the image loss of the regions in the HDR image according to the original HDR image and the reconstructed HDR image comprises the following steps:

[0148] Step 1: Perform luminance conversion on the original HDR image to obtain a first image. The first image mainly reflects the luminance value of each region in the original HDR image. In the embodiments of the present application, there are many specific algorithms for luminance conversion. For ease of understanding, examples are given below. For example, the luminance L of each region in the first image is equal to a*R+b*G+c*B. Taking different color gamut HDR images as examples, the luminance of each region (e.g. pixel) in the first image can be calculated as follows:

[0149] sRGB(R, G, B): 0.2127*R+0.7151*G+0.0721*B Formula (1-2)

[0150] P3(R, G, B): 0.2290*R+0.6917*G+0.0793*G Formula (1-3)

[0151] Rec2020(R, G, B): 0.2627*R+0.6780*G+0.0593*B Formula (1-4)

[0152] Wherein, sRGB(R, G, B), P3(R, G, B), Rec2020(R, G, B) represent three different color gamuts respectively, R is the red value of a pixel, G is the green value of a pixel, and B is the blue value of a pixel.

[0153] Step 2: Perform luminance conversion on the reconstructed HDR image to obtain a second image; the second image highlights the luminance values of each region in the reconstructed HDR image. The principle of luminance conversion is the same as step 1, and the generation order of the first image and the second image can be synchronous or sequential, which is not limited here.

[0154] Step 3: Then determine the image luminance loss of the multiple regions in the HDR image according to the first image and the second image.

[0155] It should be noted that the original HDR image, SDR image, reconstructed HDR image, first image, second image, and subsequently mentioned reference image all have multiple regions, such as 1024 regions. If the 1024 regions are sorted according to the same rule, the relative position of the i-th region in image 1 is the same as the absolute position of the i-th region in image 2, i takes any value from 1 to 1024, and image 1 and image 2 refer to any two of the original HDR image, SDR image, reconstructed HDR image, first image, second image, and subsequently mentioned reference image, i.e., these images all have multiple regions with the same distribution rule. Therefore, when determining the luminance loss, the same regions of the first image and the second image are compared. Since there are multiple regions, there are multiple comparison results, i.e., multiple image luminance losses corresponding to multiple regions.

[0156] Case two, determining the image loss of the multiple regions in the HDR image according to the original HDR image and the reconstructed HDR image includes the following steps, and Fig. 5 is a corresponding flowchart:

[0157] Step 0: Perform pixel value transformation on the original HDR image to obtain a reference image.

[0158] For example, one or more of the following operations (which can be referred to as "adjacent derivation" operations) can be performed during the pixel value transformation process:

[0159] Operation 1: Increase the R value of each pixel point in the original HDR image by a first offset Δ1 to obtain a reference image;

[0160] Operation 2: Increase the G value of each pixel point in the original HDR image by a second offset Δ2 to obtain a reference image;

[0161] Operation 3: Increase the B value of each pixel point in the original HDR image by a third offset Δ3 to obtain a reference image;

[0162] Operation 4: subtracting the first offset Δ1 from the R value of each pixel in the original HDR image to obtain a reference image;

[0163] Operation 5: subtracting the second offset Δ2 from the G value of each pixel in the original HDR image to obtain a reference image;

[0164] Operation 6: subtracting the third offset Δ3 from the B value of each pixel in the original HDR image to obtain a reference image.

[0165] Therefore, in this case, one reference image or multiple reference images can be obtained. The first offset Δ1, the second offset Δ2, and the third offset Δ3 can be three different offsets or the same offset. Alternatively, the first offset Δ1, the second offset Δ2, and the third offset Δ3 can be the minimum value unit of the R value, the G value, and the B value, respectively, or an integer multiple of the minimum value unit. The specific value can be determined according to the actual scene requirement, and is not limited herein.

[0166] As shown in FIG. 6, all the above six operations are performed, and six reference images are obtained accordingly.

[0167] As shown in FIG. 7, operations 2 and 5 are performed, and two reference images are obtained accordingly.

[0168] As shown in FIG. 8, none of the above six operations is performed, and no reference image is obtained, which corresponds to the above case 1.

[0169] It should be noted that when multiple reference images are generated, steps 1-3 need to be performed once for each reference image, and therefore the image brightness loss of multiple regions in the reconstructed HDR image is obtained for each reference image.

[0170] Step 1: performing brightness conversion processing on the reference image to obtain a first image.

[0171] Specifically, refer to step 1 in case 1, except that step 1 in case 1 extracts the brightness value in the original HDR image to obtain the first image, while this step extracts the brightness value in the reference image to obtain the first image. The principles of the two are the same, and will not be repeated here.

[0172] Step 2: performing brightness conversion processing on the reconstructed HDR image to obtain a second image.

[0173] Specifically, refer to step 2 in case 1.

[0174] Step 3: determining image brightness loss of a plurality of regions in the HDR image according to the first image and the second image.

[0175] Specifically, refer to step 3 in case 1.

[0176] It should be noted that how to use the image loss after determining the image loss in the embodiments of the present application is not limited here, which can be used to adjust the parameters of the related model, or determine some information required by the corresponding application scenario.

[0177] Optionally, the embodiments of the present application can further include one or more of the following steps S204-S208.

[0178] Step S204: determining regions meeting the loss threshold and regions not meeting the loss threshold in the plurality of regions according to the image loss of the plurality of regions.

[0179] The loss threshold is the minimum value of the estimated image brightness that can be perceived by the naked eye, and the estimation algorithm is not limited here. The value predicted by the estimation algorithm will have errors, but even with errors, it will be closer to the actual minimum value of the image brightness that can be perceived by the naked eye.

[0180] Further, after determining which regions meet the loss threshold and which regions do not meet the loss threshold according to the loss threshold, the result can be used as a basis for performing some operations, such as optimizing the related algorithm and model for reconstructing the HDR image, or for evaluating the reconstruction effect of the reconstructed HDR image, etc. In general, what the result is used for is not limited here.

[0181] As shown in FIG. 9, a flowchart for determining whether the image loss corresponding to each region meets the loss threshold is shown.

[0182] Step S205: generating the loss threshold according to the information in the original HDR image.

[0183] Specifically, the loss threshold can be generated according to the information in the original HDR image, that is, the loss threshold is generated as shown in FIG. 9.

[0184] For example, scheme one: input the parameter information in the original HDR image into Barten's model to generate the loss threshold, wherein the parameter information includes one or more of neural noise, lateral inhibition, photon noise, external noise, limited integration capacity, optical modulation transfer function, direction and time filtering. For different HDR images, the human eye perception may be different, therefore the embodiments of the present application generate targeted loss threshold for specific HDR images, it can be understood that for different original HDR images, the loss threshold generated for them is usually different. It can be understood that through the Barten's model, many factors of human eye contrast perception are considered, which is equivalent to adjusting according to the actual scene, and the loss threshold obtained has higher accuracy.

[0185] For ease of understanding, an algorithm for calculating the loss threshold C is exemplarily provided below:

[0186] C = 1 / CSF(X0, u, L) x 2 x 1 / 0.64 Formula (2-1)

[0187] k = 3.0 Formula (2-7)

[0188] σ0= 0.5 arc min Formula (2-8)

[0189] C ab = 0.08 arc min / mm Formula (2-9)

[0190] T = 0.1 sec Formula (2-10)

[0191] X max = 12° Formula (2-11)

[0192] N max = 15 cycles Formula (2-12)

[0193] η = 0.03 Formula (2-13)

[0194] Φ0= 3 x 10 -8 sec deg 2 Formula (2-14)

[0195] u0= 7 cycles / deg Formula (2-15)

[0196] p = 1.2 x 10 6 photons / sec / deg 2 / Td Formula (2-16)

[0197] In the above formulas, arc min is an angular minute, arc min / mm is an angular minute per millimeter, sec is a second, cycles is a cycle, sec deg 2 is a second square degree, cycles / deg is a cycle per degree, photons / sec / deg 2 / Td is a number of photons per second per square degree per lux.

[0198] In the above formulas, σ is a radial standard deviation of an optical point distribution function of a human eye, M opt is a contrast equation established in a spatial frequency, d is a pupil diameter of a human eye, E represents a luminance number proportional to a retinal illuminance, L is an incident luminance of an image, u is a most sensitive spatial frequency searched according to the incident luminance L, k is a signal-to-noise ratio, Sigma_0 is a small pupil size value, C_ab is a lens aberration constant, Cab is a lens aberration constant, X_max is a maximum angular size of an eye, in a degree of an integral region in an x direction, N_max is a maximum number of cycles that an eye can integrate, on which the eye can integrate information, n is a quantum detection efficiency of an eye, phi_0 is a spectral density of neural noise, u_0 is a spatial frequency threshold, a spatial frequency limit of a lateral inhibition process, beyond which the lateral inhibition stops; p is a photon conversion factor, inside which this depends on a light source; X_0 is an x-direction angular size of an object, in a degree of an object in an x direction, X_0 = 60 degrees.

[0199] Optionally, in addition to using information in the original HDR image, personalized information can also be used in generating the loss threshold, as follows: input personalized information and parameter information in the original HDR image into the Bartin model to generate the loss threshold, wherein the personalized information includes one or more of age, gender, type, and vision of a user group of the original HDR image. It can be understood that the perception of different user groups to images and their brightness is usually different, so the main user group in the application scenario of the present application can be determined first, and then the age, gender, type, and vision of the main user group are used as the consideration factors for generating the loss threshold, so that the loss threshold generated can better meet the personalized needs of the main user group.

[0200] For another example, scheme two: use square root uniform to generate (or calculate) the loss threshold in the dark of the original HDR image, and use logarithmic uniform to generate the loss threshold in the bright of the original HDR image. For example, use square root model (DeVires-Rose) to generate the loss threshold in the dark of the original HDR image, and use logarithmic uniform model (Weber-Fechner) to generate the loss threshold in the bright of the original HDR image. It can be understood that according to the different visual characteristics of the human eye in the dark and the bright, the model can basically describe the contrast perception of the human eye in this scheme, which has faster calculation speed and better practicability.

[0201] For the convenience of understanding, an algorithm for calculating the loss threshold C is exemplarily provided as follows:

[0202] Wherein, L is the brightness of a region (such as a pixel point) of the original HDR image, and the corresponding calculated loss threshold C is also the loss threshold corresponding to the region (such as a pixel point).

[0203] As shown in FIG. 10, the relationship between the loss threshold C and the image brightness L in scheme one and scheme two is illustrated.

[0204] Step S206: generate an image loss diagram.

[0205] The loss diagram includes a plurality of regions having the same distribution rule as the original HDR image, such as the same size. In the plurality of regions of the loss diagram, the luminance value of a region satisfying the loss threshold is marked as a first value, and the luminance value of a region not satisfying the loss threshold is marked as a second value. In this way, the user can visually see from the loss diagram which regions of the reconstructed HDR image satisfy the requirement of image loss (such as image luminance loss) and which regions of the reconstructed HDR image do not satisfy the requirement of image loss (such as image luminance loss). As shown in FIG. 11, the black part satisfies the requirement, and the white part does not satisfy the requirement. FIG. 11 is only one optional schematic mode, and other schematic modes can also be used.

[0206] Step S207: If the regions of the plurality of regions not satisfying the loss threshold exceed a preset proportion, adjust the model parameters.

[0207] The model parameters include parameters required for generating the reconstructed HDR image according to the HDR image and the SDR image.

[0208] Specifically, the proportion threshold can be 0 or a value greater than 0. For example, when the proportion threshold is 0, it means that the model parameters need to be adjusted as long as there is a region of the plurality of regions whose image loss does not satisfy the loss threshold (i.e., is not less than the loss threshold). For another example, when the proportion threshold is 0.5, it means that the model parameters need to be adjusted when half of the plurality of regions have image loss not satisfying the loss threshold (i.e., not less than the loss threshold). The size of the proportion threshold can be configured according to the actual scene, which is not limited here.

[0209] Regarding the adjustment of the model parameters, the following examples are given:

[0210] After analyzing the HDR image reconstruction process and the causes of the loss, the following information is obtained:

[0211] Loss root cause: The HDR reconstruction loss mainly comes from the process of saving the high-precision floating-point or integer gain map Gainmap into a low-precision fixed-point integer gain map Gainmap.

[0212] Loss example:

[0213] Gain map Gainmap floating-point to integer: float 32=>int 8;

[0214] Gain map Gainmap integer to integer: int 16=>int 8;

[0215] Loss use: after detecting loss, the reconstructed HDR image has certain quality problems compared with the original HDR image, which will cause the user to have a bad experience during browsing, so if the loss is detected or the proportion of the regions that do not meet the loss threshold in the plurality of regions exceeds the preset proportion, the model parameters are adjusted to regenerate or calculate the data of the gain map, so as to reduce or avoid loss, for example:

[0216] Case 1: increase the picture bit width of the gain map, such as increasing the picture bit width from 8 bits to 10 bits or higher, to avoid loss during the conversion process.

[0217] Case 2: reduce the gain range of the gain map, rescale the gain map to make the gain range of the gain map closer to 1, close to the effect of the SDR picture, which can reduce loss, for example, reconstructed HDR image = SDR * Gainmap (float, > 0), where the closer the gainmap is to 1, the closer the reconstructed HDR image is to the SDR image, and thus the closer the original HDR image, so the loss is smaller. Introduce a scaling factor a less than 1, reconstructed HDR image = a * SDR * Gainmap, and by reducing the value of a, the loss can be reduced.

[0218] Step S208: according to the model parameters, return to execute the step of generating a reconstructed HDR image according to the HDR image and the SDR image.

[0219] Specifically, after optimizing the model parameters, the previous step S202 is re-executed according to the optimized model parameters.

[0220] In the method described in FIG. 2, by a series of processes on the original HDR image, the SDR image and the reconstructed HDR image are obtained in turn, and then the image loss is generated according to the original HDR image and the reconstructed HDR image. The image loss can be used as feedback to optimize and adjust the image processing process, or the image loss can be used to evaluate the reconstruction process, such as identifying the visible discontinuity in the reconstructed HDR image according to the image loss. Therefore, with the image loss, it is beneficial to determine the image state or improve the image processing effect.

[0221] The above describes the method of the embodiments of the present application in detail, and the device of the embodiments of the present application is provided below.

[0222] Please refer to FIG. 12, which is a structural schematic diagram of an image loss determination apparatus provided in an embodiment of the present application. The image loss determination apparatus 120 can be the electronic device 10 mentioned above, or a component or module in the electronic device 10. The image loss determination apparatus 120 can include a first generation unit 1201, a second generation unit 1202, and a first determination unit 1203. Details of each unit are as follows.

[0223] The first generation unit 1201 is configured to generate a standard dynamic range (SDR) image from an original high dynamic range (HDR) image.

[0224] The second generation unit 1202 is configured to generate a reconstructed HDR image from the original HDR image and the SDR image.

[0225] The first determination unit 1203 is configured to determine image loss of a plurality of regions in the reconstructed HDR image according to the original HDR image and the reconstructed HDR image.

[0226] With this method, the SDR image and the reconstructed HDR image are obtained in sequence by performing a series of processes on the original HDR image. Then, the image loss is generated according to the original HDR image and the reconstructed HDR image. The image loss can be used as feedback to optimize and adjust the image processing process, or to evaluate the reconstruction process according to the image loss, such as identifying visible discontinuities in the reconstructed HDR image according to the image loss. Therefore, the image loss is helpful to determine the image state or improve the image processing effect.

[0227] In a possible implementation, the plurality of regions includes a plurality of pixels.

[0228] In another possible implementation, the apparatus further includes:

[0229] A second determination unit is configured to determine regions that meet a loss threshold and regions that do not meet the loss threshold from the plurality of regions according to the image loss of the plurality of regions, where the loss threshold is a minimum value of an image brightness that can be perceived by the naked eye.

[0230] That is, the image loss can be used to determine which regions in the plurality of regions have acceptable reconstruction loss and which regions do not have acceptable reconstruction loss, and the image loss can reflect the overall reconstruction effect of the entire image, which is helpful for users to properly optimize the subsequent image processing process.

[0231] In another possible implementation, the apparatus further includes:

[0232] A third generation unit is configured to generate the loss threshold according to information in the original HDR image.

[0233] That is to say, the loss threshold for measuring the loss of the image is determined according to the original HDR image, that is, for each HDR image, a loss threshold is generated specifically, rather than a fixed loss threshold. The threshold calculated in this way is more targeted when measuring the reconstruction effect of the reconstructed HDR image, and therefore the judgment result is more accurate.

[0234] In another possible implementation, in the aspect of generating the loss threshold according to information in the original HDR image, the third generation unit is specifically configured to:

[0235] input parameter information in the original HDR image into a Barten model to generate the loss threshold, wherein the parameter information includes one or more of neural noise, lateral inhibition, photon noise, external noise, limited integration capacity, optical modulation transfer function, direction, and temporal filtering.

[0236] It can be understood that, by using the Barten model, many factors of the contrast perception of the human eye are considered, which is equivalent to adjusting according to the actual scene, and the loss threshold obtained has higher accuracy.

[0237] In another possible implementation, in the aspect of inputting the parameter information in the original HDR image into the Barten model to generate the loss threshold, the third generation unit is specifically configured to:

[0238] input personalized information and the parameter information in the original HDR image into the Barten model to generate the loss threshold, wherein the personalized information includes one or more of the age, gender, type, and vision of a user group of the original HDR image.

[0239] It can be understood that, in this implementation, the generation of the loss threshold not only considers the characteristics of the original HDR image, but also considers the perception characteristics of the user group of the HDR image to the image information, and therefore the loss threshold generated has higher accuracy and better effect when measuring whether the image loss can be accepted by the user group.

[0240] In another possible implementation, in the aspect of generating the loss threshold according to information in the original HDR image, the third generation unit is specifically configured to:

[0241] generate the loss threshold using square root uniformity in dark places of the original HDR image and using logarithmic uniformity in bright places of the original HDR image by using the Schreiber model.

[0242] It can be understood that in this scheme, according to different visual characteristics of human eyes in dark and light, the model can basically describe the contrast perception of human eyes, and has faster calculation speed and better practicability.

[0243] In yet another possible implementation, the apparatus further includes:

[0244] A fourth generating unit is configured to generate an image loss schematic diagram, wherein the loss schematic diagram includes a plurality of regions having the same distribution rule as the original HDR image, and in the plurality of regions of the loss schematic diagram, a luminance value of a region satisfying a loss threshold is marked as a first value, and a luminance value of a region not satisfying the loss threshold is marked as a second value.

[0245] It can be understood that the user or the developer can directly view the regions satisfying the image loss condition and the regions not satisfying the image loss condition through the loss schematic diagram, which is beneficial to the user to quickly and accurately perform subsequent related operations.

[0246] In yet another possible implementation, in terms of determining the image loss of a plurality of regions in the HDR image according to the original HDR image and the reconstructed HDR image, the first determining unit is specifically configured to:

[0247] perform luminance conversion processing on the original HDR image to obtain a first image;

[0248] perform luminance conversion processing on the reconstructed HDR image to obtain a second image;

[0249] determine the image luminance loss of the plurality of regions in the HDR image according to the first image and the second image.

[0250] In yet another possible implementation, in terms of determining the image loss of a plurality of regions in the HDR image according to the original HDR image and the reconstructed HDR image, the first determining unit is specifically configured to:

[0251] perform pixel value transformation processing on the original HDR image to obtain a reference image;

[0252] perform luminance conversion processing on the reference image to obtain a first image;

[0253] perform luminance conversion processing on the reconstructed HDR image to obtain a second image;

[0254] determine the image luminance loss of the plurality of regions in the HDR image according to the first image and the second image.

[0255] In this way, the original HDR image is first subjected to pixel value transformation to obtain a reference image; subsequently, the image loss between the reconstructed HDR image and the reference image is determined, which is better to determine whether the transition appears discontinuity when the brightness changes.

[0256] In another possible implementation, in the aspect of subjecting the original HDR image to pixel value transformation to obtain a reference image, the first determining unit is specifically configured to:

[0257] perform one or more of the following operations:

[0258] increase the R value of each pixel point in the original HDR image by a first offset to obtain a reference image;

[0259] increase the G value of each pixel point in the original HDR image by a second offset to obtain a reference image;

[0260] increase the B value of each pixel point in the original HDR image by a third offset to obtain a reference image;

[0261] subtract the R value of each pixel point in the original HDR image by a first offset to obtain a reference image;

[0262] subtract the G value of each pixel point in the original HDR image by a second offset to obtain a reference image;

[0263] subtract the B value of each pixel point in the original HDR image by a third offset to obtain a reference image.

[0264] In another possible implementation, the apparatus further includes:

[0265] an adjusting unit configured to adjust a model parameter in a case where a region that does not satisfy a loss threshold in the plurality of regions exceeds a preset proportion, wherein the model parameter includes a parameter required for generating a reconstructed HDR image from the HDR image and the SDR image;

[0266] trigger the second generating unit to return to perform the step of generating a reconstructed HDR image from the HDR image and the SDR image according to the model parameter.

[0267] It can be understood that taking the image loss as the basis for adjusting the model parameter is beneficial to subsequently reconstructing a better HDR image.

[0268] In another possible implementation, in the aspect of generating a reconstructed HDR image from the original HDR image and the SDR image, the second generating unit is specifically configured to:

[0269] generate a gain map according to the original HDR image and the SDR image;

[0270] generate a reconstructed HDR image according to the gain map and the SDR image.

[0271] It can be seen that the gain map is generated in the process of reconstructing the HDR image, and the gain map can be used to enhance the dynamic range of the SDR image, so as to obtain the effect of the HDR image.

[0272] It should be noted that the implementation of each unit can also correspond to the description of the corresponding method embodiment shown in FIG. 2.

[0273] The embodiment of the present application also provides a chip system, which comprises at least one processor, a memory and an interface circuit, the memory, the transceiver and the at least one processor are interconnected through a line, and the at least one memory stores a computer program; when the computer program is executed by the processor, the method process shown in FIG. 2 is realized.

[0274] The embodiment of the present application also provides a computer readable storage medium, which stores a computer program, and when the computer program is run on a processor, the method process shown in FIG. 2 is realized.

[0275] The embodiment of the present application also provides a computer program product, and when the computer program product is run on a processor, the method process shown in FIG. 2 is realized.

[0276] In summary, through the embodiment of the present application, through a series of processing on the original HDR image, the SDR image and the reconstructed HDR image are obtained in turn, and then the image loss is generated according to the original HDR image and the reconstructed HDR image. The image loss can be used as feedback to optimize and adjust the image processing process, or the image loss can be used to evaluate the reconstruction process, such as identifying the visible fault in the reconstructed HDR image according to the image loss. Therefore, with the image loss, it is beneficial to determine the image state or improve the image processing effect.

[0277] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by a computer program related hardware, the computer program can be stored in a computer readable storage medium, and the computer program can include the processes of the above-mentioned method embodiments when executed. The storage medium includes ROM or random storage memory RAM, magnetic disc or optical disc and various storage computer program code media.

Claims

1. An image loss determination method, characterized in that, The method comprises: generating a standard dynamic range (SDR) image from an original high dynamic range (HDR) image; generating a reconstructed HDR image from the original HDR image and the SDR image; determining image loss of a plurality of regions in the reconstructed HDR image according to the original HDR image and the reconstructed HDR image.

2. The method of claim 1, wherein, The plurality of regions comprises a plurality of pixels.

3. The method according to claim 1 or 2, characterized in that, The method further comprises: determining regions satisfying a loss threshold and regions not satisfying the loss threshold in the plurality of regions according to the image loss of the plurality of regions, wherein the loss threshold is a minimum value of an image brightness that can be perceived by the naked eye.

4. The method according to any one of claims 1 to 3, characterized in that, The method further comprises: generating the loss threshold according to information in the original HDR image.

5. The method of claim 4, wherein, The generating the loss threshold according to information in the original HDR image comprises: inputting parameter information in the original HDR image into a Bartlett model to generate the loss threshold, wherein the parameter information comprises one or more of neural noise, lateral inhibition, photon noise, external noise, limited integration capacity, optical modulation transfer function, orientation, and temporal filtering.

6. The method of claim 5, wherein, The inputting the parameter information in the original HDR image into the Bartlett model to generate the loss threshold comprises: inputting personalized information and the parameter information in the original HDR image into the Bartlett model to generate the loss threshold, wherein the personalized information comprises one or more of age, gender, type, and vision of a user group of the original HDR image.

7. The method of claim 4, wherein, The generating the loss threshold according to information in the original HDR image comprises: generating a loss threshold using square root uniformity in dark places of the original HDR image and using logarithmic uniformity in bright places of the original HDR image through a Schreiber model.

8. The method of claim 3, wherein, The method further comprises: generating an image loss diagram, wherein the loss diagram comprises a plurality of regions having the same distribution rule as the original HDR image, and in the plurality of regions of the loss diagram, a brightness value of a region satisfying the loss threshold is marked as a first value, and a brightness value of a region not satisfying the loss threshold is marked as a second value.

9. The method according to any one of claims 1 to 8, characterized in that, The determining the image loss of the plurality of regions in the HDR image according to the original HDR image and the reconstructed HDR image comprises: performing brightness conversion processing on the original HDR image to obtain a first image; performing brightness conversion processing on the reconstructed HDR image to obtain a second image; determining image brightness loss of the plurality of regions in the HDR image according to the first image and the second image.

10. The method according to any one of claims 1 to 8, characterized in that, The determining the image loss of the plurality of regions in the HDR image according to the original HDR image and the reconstructed HDR image comprises: performing pixel value transformation processing on the original HDR image to obtain a reference image; performing brightness conversion processing on the reference image to obtain a first image; performing brightness conversion processing on the reconstructed HDR image to obtain a second image; determining image brightness loss of the plurality of regions in the HDR image according to the first image and the second image.

11. The method of claim 10, wherein, The pixel value transformation processing on the original HDR image obtains a reference image, including: One or more of the following operations are performed: The R value of each pixel point in the original HDR image is increased by a first offset to obtain a reference image; The G value of each pixel point in the original HDR image is increased by a second offset to obtain a reference image; The B value of each pixel point in the original HDR image is increased by a third offset to obtain a reference image; The R value of each pixel point in the original HDR image is reduced by a first offset to obtain a reference image; The G value of each pixel point in the original HDR image is reduced by a second offset to obtain a reference image; The B value of each pixel point in the original HDR image is reduced by a third offset to obtain a reference image.

12. The method of claim 3, wherein, Further comprising: If the regions that do not meet the loss threshold in the plurality of regions exceed a preset proportion, adjusting the model parameters, wherein the model parameters include parameters required for generating a reconstructed HDR image from the HDR image and the SDR image; Returning to the step of generating a reconstructed HDR image from the HDR image and the SDR image according to the model parameters.

13. The method according to any one of claims 1 to 12, characterized in that, The generating a reconstructed HDR image from the original HDR image and the SDR image includes: Generating a gain map from the original HDR image and the SDR image; Generating a reconstructed HDR image from the gain map and the SDR image.

14. An electronic device, comprising: The electronic device includes one or more processors; a memory; wherein the memory is coupled with the one or more processors, the memory is used to store a computer program, and the one or more processors invoke the computer program to make the electronic device execute the method of any one of claims 1-13.

15. A chip system, characterized by The chip system is applied to an electronic device, and the chip system includes one or more processors, and the processor is used to invoke a computer program to make the electronic device execute the method of any one of claims 1-13.

16. A computer program product comprising instructions, characterized in that, When the computer program product runs on an electronic device, it makes the electronic device execute the method of any one of claims 1-13.

17. A computer readable storage medium comprising a computer program, characterized in that, When the computer program runs on an electronic device, it makes the electronic device execute the method of any one of claims 1-13.

Images