Image enhancement method and system based on image processing chip

Abstract

The invention relates to an image enhancement method and system based on an image processing chip, and relates to the field of general image data processing or generation. The method comprises the steps that an AI prediction model is adopted to intelligently predict the signal-to-noise ratio of an enhanced image obtained after an image to be enhanced is subjected to image enhancement processing of a target image enhancement algorithm based on various collection information of an image processing chip, multiple analysis parameters of the image to be enhanced and algorithm data corresponding to the target image enhancement algorithm; and selecting an optimal image enhancement algorithm of the to-be-enhanced image based on the signal-to-noise ratio of each part corresponding to each image enhancement algorithm. The invention also relates to an image enhancement system based on the image processing chip. According to the invention, aiming at the technical problem that respective optimal image enhancement algorithms of different images cannot be predicted, the AI prediction model with a customized structure can be adopted to predict the image enhancement effect obtained by the image processing chip by adopting any type of image enhancement mode, so that the technical problem is solved.

Description

Technical Field

[0001] The present invention relates to the general field of image data processing or generation, and in particular to an image enhancement method and system based on an image processing chip. Background Art

[0002] Usually, image processing chips are used to perform general image data processing or generation. Image processing chips are the core of graphics cards, and the main technical specifications and performance of graphics cards basically depend on the technical type and performance of image processing chips. The technical advancement of image processing chips is mainly measured by their 2D / 3D graphics processing capabilities, the data bit width of the chip graphics processing engine, the data bus width between the display memory and the supported video memory type capacity, the working clock frequency of the internal RAMDAC (analog / digital converter), the number of pixel rendering processing pipelines, the types of graphics application program interfaces (APIs) supported, and the chip production process technology level.

[0003] For example, Chinese invention patent publication CN113014822A proposes an image enhancement processing system and method for a camera image processing chip, the system includes a shooting module, an information transmission module, a connection control module, a parsing module, and an image enhancement module. The shooting module records the scene to be shot, and the shooting module transmits the information to the connection terminal through the information transmission module and saves it. The staff can choose to crop and save the required part of the image through the terminal control module, and use the parsing module to decompose the shot picture into images, and use the image enhancement module to enhance and save the decomposed image, so as to effectively enhance the image, increase the shooting information, and save the information in an all-round way for later selection and image enhancement.

[0004] For example, the Chinese invention patent publication CN117437163A proposes an image enhancement method, an image processing chip and an image enhancement video system, wherein the image enhancement method includes: using calibrated images for classification training to establish a neural network processor; optimizing and adjusting the parameters of each category of images to obtain the best image quality, and saving the corresponding optimal parameter values ​​in the neural network processor; using the neural network processor to classify each input frame of the image, and calling the optimal parameter values ​​of the corresponding category to optimize the image; and performing partition calculations on each input frame of the image to obtain the backlight parameters of each partition. The AI ​​algorithm is used to calculate each frame of the input video and optimize the image quality of each frame in real time, effectively improving the image quality of the video displayed on the screen.

[0005] It can be seen that in the above-mentioned prior art, the same image enhancement mode is used to perform image enhancement processing with the same algorithm on various picture contents. However, since different picture contents collected in real time may correspond to different optimal image enhancement algorithms, if the same type of image enhancement algorithm is used to perform the same image enhancement processing on different picture contents, it is obvious that it cannot guarantee that each frame of the picture content will achieve the best image enhancement effect. Therefore, an image enhancement mechanism is needed that can select different types of image enhancement algorithms that are more suitable for different picture contents when the image processing chip performs image enhancement processing, so as to obtain the best image enhancement effect for each. Summary of the invention

[0006] In order to solve the technical defects in the prior art, the present invention provides an image enhancement method and system based on an image processing chip, which can, for the same image, use a customized AI prediction model to predict the image enhancement effect obtained by the image processing chip using any type of image enhancement mode, while selectively selecting multiple analytical parameters of a reference image, image data of the reference image, various acquisition information of the image processing chip, multiple analytical parameters of the same image, and algorithm data corresponding to the image enhancement algorithm used as prediction basic data, and then compare various picture enhancement effects to select the image enhancement mode with the best picture enhancement effect and actually put it into the image processing chip for enhancement processing, thereby using an artificial intelligence mechanism to improve the enhancement performance of the image processing chip.

[0007] According to a first aspect of the present invention, there is provided an image enhancement method based on an image processing chip, which runs on the image processing chip, and the method comprises: Collecting the memory capacity, operating frequency, fastest memory read and write speed, chip weight and chip volume of the image processing chip and outputting them as various collected information of the image processing chip; Analyze the signal-to-noise ratio, the number of high-amplitude noise types, the horizontal resolution, the vertical resolution, the pixel value redundancy, and the number of main objects of the image to be enhanced to output as multiple analytical parameters of the image to be enhanced; Enter the algorithm data corresponding to each image enhancement algorithm. The algorithm data corresponding to each image enhancement algorithm is the code length, type identification and processed reference image data after image enhancement processing by the image enhancement algorithm. Performing multiple learning on the feedforward neural network to obtain a feedforward neural network after performing the multiple learning and outputting it as an AI prediction model, wherein the number of learning times of the multiple learning is monotonically positively correlated with the pixel points of the reference image; The AI ​​prediction model is used to intelligently predict the signal-to-noise ratio of an enhanced processed image obtained after the image to be enhanced is subjected to image enhancement processing by the target image enhancement algorithm based on multiple analytical parameters of the reference image, image data of the reference image, various acquisition information of the image processing chip, multiple analytical parameters of the image to be enhanced, and algorithm data corresponding to the target image enhancement algorithm; Selecting the optimal image enhancement algorithm for the image to be enhanced based on the signal-to-noise ratios corresponding to the various image enhancement algorithms for execution by the image processing chip; Among them, selecting the optimal image enhancement algorithm of the image to be enhanced based on the signal-to-noise ratios corresponding to the various image enhancement algorithms for execution by the image processing chip includes: the various image enhancement processing corresponding to the various image enhancement algorithms are histogram correction, enhancement processing based on linear grayscale transformation, enhancement processing based on piecewise linear grayscale transformation, enhancement processing based on logarithmic transformation, enhancement processing based on exponential transformation and frequency domain enhancement processing, and the optimal image enhancement algorithm of the image to be enhanced is the image enhancement algorithm corresponding to the signal-to-noise ratio with the largest value among the various signal-to-noise ratios.

[0008] According to a second aspect of the present invention, there is provided an image enhancement system based on an image processing chip, which runs on the image processing chip, and the system comprises: A chip acquisition device, used to acquire the memory capacity, operating frequency, fastest memory read and write speed, chip weight and chip volume of the image processing chip and output it as various acquisition information of the image processing chip; A content analysis device, used for analyzing the signal-to-noise ratio, the number of high-amplitude noise types, the horizontal resolution, the vertical resolution, the pixel value redundancy and the number of main objects of the image to be enhanced to output as multiple analysis parameters of the image to be enhanced; A data entry device, used to enter various algorithm data corresponding to various image enhancement algorithms, wherein the single algorithm data corresponding to each image enhancement algorithm is the code length and type identification of the image enhancement algorithm and the processed reference image data after image enhancement processing by the image enhancement algorithm; A model building device, used to perform multiple learning on the feedforward neural network to obtain a feedforward neural network after performing multiple learning and output it as an AI prediction model, wherein the number of learning times of the multiple learning is monotonically positively correlated with the pixel points of the reference image; An effect prediction device is connected to the chip acquisition device, the content analysis device, the data entry device and the model building device respectively, and is used to use the AI ​​prediction model to intelligently predict the signal-to-noise ratio of the enhanced processed image obtained after the image to be enhanced is subjected to image enhancement processing by the target image enhancement algorithm based on multiple analysis parameters of the reference image, image data of the reference image, various acquisition information of the image processing chip, multiple analysis parameters of the image to be enhanced and algorithm data corresponding to the target image enhancement algorithm; an algorithm selection device, connected to the effect prediction device, for selecting the optimal image enhancement algorithm of the image to be enhanced based on the signal-to-noise ratios corresponding to the various image enhancement algorithms for execution by the image processing chip; Among them, selecting the optimal image enhancement algorithm of the image to be enhanced based on the signal-to-noise ratios corresponding to the various image enhancement algorithms for execution by the image processing chip includes: the various image enhancement processing corresponding to the various image enhancement algorithms are histogram correction, enhancement processing based on linear grayscale transformation, enhancement processing based on piecewise linear grayscale transformation, enhancement processing based on logarithmic transformation, enhancement processing based on exponential transformation and frequency domain enhancement processing, and the optimal image enhancement algorithm of the image to be enhanced is the image enhancement algorithm corresponding to the signal-to-noise ratio with the largest value among the various signal-to-noise ratios.

[0009] It can be seen that the present invention has at least the following four key invention points: Invention point A: For an image processing scenario in which the same image processing chip enhances the same picture content, an artificial intelligence model is used to respectively predict various picture enhancement effects obtained after different types of image enhancement modes enhance the same picture content, and various picture enhancement effects are compared to select the image enhancement mode with the best picture enhancement effect to actually be used in the enhancement processing operation of the image processing chip on the same picture content, thereby optimizing the image enhancement effect of the image processing chip on different picture contents. Specifically, different types of image enhancement modes include histogram correction, enhancement processing based on linear grayscale transformation, enhancement processing based on piecewise linear grayscale transformation, enhancement processing based on logarithmic transformation, enhancement processing based on exponential transformation, and frequency domain enhancement processing; Invention point B: A customized AI prediction model is used to predict the image enhancement effect obtained by an image processing chip using any type of image enhancement mode. The AI ​​prediction model is a feedforward neural network that has performed multiple learnings, and the number of learnings in the multiple learnings is monotonically positively correlated with the pixels of the reference image, thereby ensuring the reliability and stability of the intelligent prediction results. Invention point C: multiple basic data are selected for the AI ​​prediction model to perform intelligent prediction of image enhancement effects. The multiple basic data are multiple analytical parameters of the reference image, image data of the reference image, various acquisition information of the image processing chip, multiple analytical parameters of the image to be enhanced, and algorithm data corresponding to the image enhancement algorithm used. The selection and use of the above sufficient and comprehensive basic data further ensures the reliability and stability of the intelligent prediction results; Invention point D: In each learning performed on the feedforward neural network, the known signal-to-noise ratio of an enhanced image obtained after an image is enhanced by a certain image enhancement algorithm is used as the output content of the feedforward neural network, and multiple analytical parameters of the reference image, image data of the reference image, various acquisition information of the image processing chip, multiple analytical parameters of the certain image and algorithm data corresponding to the certain image enhancement algorithm are used as multiple input contents of the feedforward neural network to perform this learning, thereby ensuring the efficiency and effect of each learning of the feedforward neural network. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The embodiments of the present invention will be described below with reference to the accompanying drawings, wherein: Figure 1 The figure is a technical flow chart of image enhancement based on an image processing chip according to the present invention.

[0011] Figure 2 The figure is a flowchart of the steps of an image enhancement method based on an image processing chip according to Embodiment 1 of the present invention.

[0012] Figure 3 The figure is a flowchart showing a method for image enhancement based on an image processing chip according to Embodiment 2 of the present invention.

[0013] Figure 4 The figure is a flowchart showing a method for image enhancement based on an image processing chip according to Embodiment 3 of the present invention.

[0014] Figure 5 1 is an internal structure diagram of an image enhancement system based on an image processing chip according to Embodiment 4 of the present invention.

[0015] Figure 6 1 is an internal structure diagram of an image enhancement system based on an image processing chip according to Embodiment 5 of the present invention.

[0016] Figure 7 1 is an internal structure diagram of an image enhancement system based on an image processing chip according to Embodiment 6 of the present invention. DETAILED DESCRIPTION

[0017] like Figure 1 As shown, a technical flow chart of image enhancement based on an image processing chip according to the present invention is given.

[0018] like Figure 1 As shown, the specific technical process of the present invention is as follows: Technical process 1: Design a customized AI prediction model to predict the image enhancement effect obtained by the image processing chip using any type of image enhancement mode; Specifically, the structural customization of the AI ​​prediction model is reflected in the following three aspects: Aspect 1: The AI ​​prediction model is a feedforward neural network after multiple learnings; Aspect 2: The learning times of the feedforward neural network are monotonically positively correlated with the pixels of the reference image, thus ensuring the reliability and stability of the intelligent prediction results; Aspect 2: In each learning performed on the feedforward neural network, a known signal-to-noise ratio of an enhanced image obtained after an image is enhanced by an image enhancement algorithm is used as the output content of the feedforward neural network, and multiple analytical parameters of a reference image, image data of the reference image, various acquisition information of an image processing chip, multiple analytical parameters of a certain image, and algorithm data corresponding to a certain image enhancement algorithm are used as multiple input contents of the feedforward neural network to perform this learning, thereby ensuring the efficiency and effect of each learning of the feedforward neural network; Technical process 2: Targeted selection of multiple basic data for AI prediction model to perform intelligent prediction of image enhancement effect; Specifically, the multiple basic data are multiple analytical parameters of the reference image, image data of the reference image, various acquisition information of the image processing chip, multiple analytical parameters of the image to be enhanced, and algorithm data corresponding to the image enhancement algorithm used; It can be seen that the selection and use of the above-mentioned sufficient and comprehensive basic data further ensure the reliability and stability of the intelligent prediction results; Technical process three: For the image processing scenario where the same image processing chip enhances the same image content, the AI ​​prediction model designed in technical process one is used to predict the various image enhancement effects obtained after different types of image enhancement modes enhance the same image content based on multiple basic data selected in technical process two; Technical process 4: for the same picture content, compare various picture enhancement effects to select an image enhancement mode with the best picture enhancement effect, and actually input the selected image enhancement mode into the image processing chip to perform an enhancement processing operation on the same picture content; It can be seen that the above-mentioned technical processes optimize the image enhancement effect of the image processing chip on different picture contents; For example, different types of image enhancement modes include histogram correction, enhancement processing based on linear grayscale transformation, enhancement processing based on piecewise linear grayscale transformation, enhancement processing based on logarithmic transformation, enhancement processing based on exponential transformation, and frequency domain enhancement processing.

[0019] The key points of the present invention are: customized structural design of AI prediction model, targeted selection of multiple basic data for intelligent prediction of image enhancement effect, and comparison and selection of the best image enhancement processing mode in the scenario where the same image processing chip is used to perform enhancement processing on the same picture content.

[0020] In this way, when an image processing chip is used for enhancement processing, different picture contents can obtain their own optimal image enhancement processing mode, thereby improving the intelligence level of the image enhancement processing of the image processing chip and the image enhancement effect.

[0021] In the following, an image enhancement method and system based on an image processing chip of the present invention will be specifically described in the form of an embodiment.

[0022] Example 1 Figure 2 The figure is a flowchart of the steps of an image enhancement method based on an image processing chip according to Embodiment 1 of the present invention.

[0023] like Figure 2 As shown, the image enhancement method based on the image processing chip runs on the image processing chip and includes the following steps: Step S11: collecting the memory capacity, operating frequency, fastest memory read and write speed, chip weight and chip volume of the image processing chip and outputting them as various collected information of the image processing chip; For example, collecting the memory capacity, operating frequency, fastest memory read / write speed, chip weight and chip volume of the image processing chip and outputting them as various collected information of the image processing chip includes: using a plurality of different parameter collection components to respectively collect the memory capacity, operating frequency, fastest memory read / write speed, chip weight and chip volume of the image processing chip, and outputting the memory capacity, operating frequency, fastest memory read / write speed, chip weight and chip volume of the image processing chip as various collected information of the image processing chip; Step S12: analyzing the signal-to-noise ratio, the number of high-amplitude noise types, the horizontal resolution, the vertical resolution, the pixel value redundancy, and the number of main objects of the image to be enhanced to output as multiple analytical parameters of the image to be enhanced; Specifically, analyzing the signal-to-noise ratio, the number of high-amplitude noise types, the horizontal resolution, the vertical resolution, the pixel value redundancy, and the number of main objects of the image to be enhanced to output as multiple analytical parameters of the image to be enhanced includes: the horizontal resolution of the image to be enhanced is the number of pixel columns of the image to be enhanced, and the vertical resolution of the image to be enhanced is the number of pixel rows of the image to be enhanced; Step S13: inputting the algorithm data corresponding to various image enhancement algorithms respectively, wherein the single algorithm data corresponding to each image enhancement algorithm is the code length, type identification and processed reference image data after image enhancement processing by the image enhancement algorithm; Specifically, various algorithm data corresponding to various image enhancement algorithms are entered, and the single algorithm data corresponding to each image enhancement algorithm is the code length, type identification and processed reference image data after image enhancement processing by the image enhancement algorithm, including: the code of each image enhancement algorithm is a computer executable instruction for implementing the corresponding type of image enhancement algorithm; Step S14: performing multiple learning on the feedforward neural network to obtain a feedforward neural network after performing multiple learning and outputting it as an AI prediction model, wherein the number of learning times of the multiple learning is monotonically positively correlated with the pixel points of the reference image; For example, multiple learning is performed on the feedforward neural network to obtain the feedforward neural network after the multiple learning is performed and output as the AI ​​prediction model, and the monotonically positive correlation between the number of learning times of the multiple learning and the pixel points of the reference image includes: when the number of pixels of the reference image is 3840×2160, the value of the number of learning times of the multiple learning is 300, when the number of pixels of the reference image is 2560×1440, the value of the number of learning times of the multiple learning is 200, when the number of pixels of the reference image is 1920×1080, the value of the number of learning times of the multiple learning is 150, when the number of pixels of the reference image is 1600×900, the value of the number of learning times of the multiple learning is 120, when the number of pixels of the reference image is 1280×720, the value of the number of learning times of the multiple learning is 100, and so on; Step S15: using the AI ​​prediction model to intelligently predict the signal-to-noise ratio of an enhanced processed image obtained after the image to be enhanced is subjected to image enhancement processing by the target image enhancement algorithm based on multiple analytical parameters of the reference image, image data of the reference image, various acquisition information of the image processing chip, multiple analytical parameters of the image to be enhanced, and algorithm data corresponding to the target image enhancement algorithm; Step S16: selecting the optimal image enhancement algorithm of the image to be enhanced based on the signal-to-noise ratios corresponding to the various image enhancement algorithms for execution by the image processing chip; Among them, selecting the optimal image enhancement algorithm of the image to be enhanced based on the signal-to-noise ratios corresponding to the various image enhancement algorithms for the image processing chip to execute includes: the various image enhancement processes corresponding to the various image enhancement algorithms are histogram correction, enhancement process based on linear grayscale transformation, enhancement process based on piecewise linear grayscale transformation, enhancement process based on logarithmic transformation, enhancement process based on exponential transformation and frequency domain enhancement process, and the optimal image enhancement algorithm of the image to be enhanced is the image enhancement algorithm corresponding to the signal-to-noise ratio with the largest value among the signal-to-noise ratios; Among them, performing multiple learning on the feedforward neural network to obtain the feedforward neural network after performing multiple learning and outputting it as the AI ​​prediction model, and the number of learning times of the multiple learning is monotonically positively correlated with the pixel points of the reference image, including: in each learning performed on the feedforward neural network, using the known signal-to-noise ratio of an enhanced processed image obtained after an image is enhanced by a certain image enhancement algorithm as the output content of the feedforward neural network, using multiple analytical parameters of the reference image, image data of the reference image, various acquisition information of the image processing chip, multiple analytical parameters of the certain image, and algorithm data corresponding to the certain image enhancement algorithm as multiple input contents of the feedforward neural network to perform this learning; Among them, the AI ​​prediction model is used to intelligently predict the signal-to-noise ratio of the enhanced processed image obtained after the image to be enhanced is subjected to image enhancement processing by the target image enhancement algorithm based on multiple analytical parameters of the reference image, image data of the reference image, various acquisition information of the image processing chip, multiple analytical parameters of the image to be enhanced, and algorithm data corresponding to the target image enhancement algorithm, including: the image data of the reference image is each pixel value corresponding to each pixel point of the reference image; For example, the image data of the reference image are pixel values ​​corresponding to the pixels of the reference image, including: the pixel values ​​corresponding to the pixels of the reference image are brightness values ​​corresponding to the pixels of the reference image; And wherein, the AI ​​prediction model is used to intelligently predict the signal-to-noise ratio of the enhanced processed image obtained after the image to be enhanced is processed by the target image enhancement algorithm based on multiple analytical parameters of the benchmark image, the image data of the benchmark image, various acquisition information of the image processing chip, multiple analytical parameters of the image to be enhanced, and the algorithm data corresponding to the target image enhancement algorithm, and also includes: the target image enhancement algorithm is histogram correction, enhancement processing based on linear grayscale transformation, enhancement processing based on piecewise linear grayscale transformation, enhancement processing based on logarithmic transformation, enhancement processing based on exponential transformation, and frequency domain enhancement processing. The algorithm corresponds to any one of the image enhancement processing.

[0024] Example 2 Figure 3 The figure is a flowchart showing a method for image enhancement based on an image processing chip according to Embodiment 2 of the present invention.

[0025] like Figure 3 As shown, after performing multiple learning on the feedforward neural network to obtain the feedforward neural network after performing multiple learning and outputting it as the AI ​​prediction model, and the number of learning times of the multiple learning is monotonically positively associated with the pixel points of the reference image, that is, after step S14, the image enhancement method based on the image processing chip also includes: Step S17: using a data storage chip to receive various model parameters of the AI ​​prediction model, and completing model storage of the AI ​​prediction model by storing various model parameters of the AI ​​prediction model; By way of example, a data storage chip is used to receive various model parameters of the AI ​​prediction model, and model storage of the AI ​​prediction model is completed by storing various model parameters of the AI ​​prediction model, including: the data storage chip is a TF storage chip, an MMC storage chip or a FLASH storage chip.

[0026] Example 3 Figure 4 The figure is a flowchart of the steps of an image enhancement method based on an image processing chip according to Embodiment 3 of the present invention.

[0027] like Figure 4 As shown, after selecting the optimal image enhancement algorithm of the image to be enhanced based on the signal-to-noise ratios corresponding to the various image enhancement algorithms for execution by the image processing chip, that is, after step S16, the image enhancement method based on the image processing chip further includes: Step S18: receiving the optimal image enhancement algorithm of the image to be enhanced, and performing on-site display of the optimal image enhancement algorithm of the image to be enhanced through the LCD display array; For example, receiving the optimal image enhancement algorithm for the image to be enhanced, and performing an on-site display of the optimal image enhancement algorithm for the image to be enhanced through an LCD display array includes: optionally using a liquid crystal display screen or an LED display array to replace the LCD display array, for performing an on-site display of the optimal image enhancement algorithm for the image to be enhanced.

[0028] In any of the above embodiments 1-3, optionally, in the image enhancement method based on an image processing chip: The signal-to-noise ratio, the number of high-amplitude noise types, the horizontal resolution, the vertical resolution, the pixel value redundancy and the number of main objects of the image to be enhanced are analyzed to output as multiple analytical parameters of the image to be enhanced, including: the high-amplitude noise type of the image to be enhanced is a noise type whose amplitude in the image to be enhanced exceeds a preset noise amplitude; Wherein, analyzing the signal-to-noise ratio, the number of high-amplitude noise types, the horizontal resolution, the vertical resolution, the pixel value redundancy and the number of main objects of the image to be enhanced as the multiple analytical parameter outputs of the image to be enhanced further includes: after the respective pixel values ​​corresponding to the respective pixel points of the image to be enhanced are deduplicated, the number of the remaining multiple pixel values ​​is a first number, the number of the respective pixel values ​​corresponding to the respective pixel points of the image to be enhanced is a second number, the first number is subtracted from the second number to obtain a third number, and the third number is divided by the second number to obtain the pixel value redundancy of the image to be enhanced; Wherein, analyzing the signal-to-noise ratio, the number of high-amplitude noise types, the horizontal resolution, the vertical resolution, the pixel value redundancy and the number of main objects of the image to be enhanced as the output of multiple analytical parameters of the image to be enhanced further includes: the main object of the image to be enhanced is an object in the image to be enhanced that occupies an area ratio of the image to be enhanced that is greater than or equal to a set ratio limit; Specifically, the main object of the image to be enhanced is an object in the image to be enhanced that occupies an area ratio of the image to be enhanced greater than or equal to a set ratio limit, and further includes: the main object of the image to be enhanced may be a relatively large human body, tree branch or house, etc.; The main object of the image to be enhanced is an object in the image to be enhanced that occupies an area ratio of the image to be enhanced that is greater than or equal to a set ratio limit, comprising: removing background pixels from pixels of the image to be enhanced to obtain foreground pixels, and performing fitting processing on the foreground pixels to obtain objects in the image to be enhanced; Among them, the algorithm data corresponding to various image enhancement algorithms are entered respectively, and the single algorithm data corresponding to each image enhancement algorithm is the code length, type identification and processed reference image data after image enhancement processing by the image enhancement algorithm, including: the code length of each image enhancement algorithm is the number of bits of binary data formed after the code of the image enhancement algorithm is converted into binary value; Among them, the algorithm data corresponding to various image enhancement algorithms are entered, and the single algorithm data corresponding to each image enhancement algorithm is the code length and type identification of the image enhancement algorithm, and the processed reference image data after the image enhancement processing of the image enhancement algorithm also includes: the processed reference image data after the image enhancement processing of each image enhancement algorithm is the pixel values ​​corresponding to each pixel point of the image obtained after the image enhancement processing of the image enhancement algorithm is performed on the reference image; Among them, performing multiple learning on the feedforward neural network to obtain the feedforward neural network after performing the multiple learning and outputting it as the AI ​​prediction model, wherein the number of learning times of the multiple learning and the monotonic positive correlation of the pixel points of the reference image further includes: using a numerical mapping function to represent the numerical mapping relationship of the monotonic positive correlation of the number of learning times of the multiple learning and the pixel points of the reference image; The method of using a numerical mapping function to represent the numerical mapping relationship of the monotonically positive correlation between the number of learning times of the multiple learning and the pixel points of the reference image comprises: in the numerical mapping function, the number of learning times of the multiple learning is the output parameter of the numerical mapping function, and the number of pixel points of the reference image is the input parameter of the numerical mapping function; Among them, using the AI ​​prediction model to intelligently predict the signal-to-noise ratio of the enhanced processed image obtained after the image to be enhanced is subjected to image enhancement processing by the target image enhancement algorithm based on multiple analytical parameters of the reference image, image data of the reference image, various acquisition information of the image processing chip, multiple analytical parameters of the image to be enhanced, and algorithm data corresponding to the target image enhancement algorithm also includes: synchronously inputting the multiple analytical parameters of the reference image, the image data of the reference image, various acquisition information of the image processing chip, multiple analytical parameters of the image to be enhanced, and algorithm data corresponding to the target image enhancement algorithm into the AI ​​prediction model; And wherein, using the AI ​​prediction model to intelligently predict the signal-to-noise ratio of the enhanced processed image obtained after the image to be enhanced is subjected to image enhancement processing by the target image enhancement algorithm based on multiple analytical parameters of the baseline image, the image data of the baseline image, various acquisition information of the image processing chip, multiple analytical parameters of the image to be enhanced, and algorithm data corresponding to the target image enhancement algorithm also includes: executing the AI ​​prediction model to obtain the signal-to-noise ratio of the enhanced processed image obtained after the image to be enhanced is subjected to image enhancement processing by the target image enhancement algorithm, which is output by the AI ​​prediction model.

[0029] Example 4 Figure 5 1 is an internal structure diagram of an image enhancement system based on an image processing chip according to Embodiment 4 of the present invention.

[0030] like Figure 5 As shown, the image enhancement system based on the image processing chip runs on the image processing chip and includes the following components: A chip acquisition device, used to acquire the memory capacity, operating frequency, fastest memory read and write speed, chip weight and chip volume of the image processing chip and output it as various acquisition information of the image processing chip; For example, collecting the memory capacity, operating frequency, fastest memory read / write speed, chip weight and chip volume of the image processing chip and outputting them as various collected information of the image processing chip includes: using a plurality of different parameter collection components to respectively collect the memory capacity, operating frequency, fastest memory read / write speed, chip weight and chip volume of the image processing chip, and outputting the memory capacity, operating frequency, fastest memory read / write speed, chip weight and chip volume of the image processing chip as various collected information of the image processing chip; A content analysis device, used for analyzing the signal-to-noise ratio, the number of high-amplitude noise types, the horizontal resolution, the vertical resolution, the pixel value redundancy and the number of main objects of the image to be enhanced to output as multiple analysis parameters of the image to be enhanced; Specifically, analyzing the signal-to-noise ratio, the number of high-amplitude noise types, the horizontal resolution, the vertical resolution, the pixel value redundancy, and the number of main objects of the image to be enhanced to output as multiple analytical parameters of the image to be enhanced includes: the horizontal resolution of the image to be enhanced is the number of pixel columns of the image to be enhanced, and the vertical resolution of the image to be enhanced is the number of pixel rows of the image to be enhanced; A data entry device, used to enter various algorithm data corresponding to various image enhancement algorithms, wherein the single algorithm data corresponding to each image enhancement algorithm is the code length and type identification of the image enhancement algorithm and the processed reference image data after image enhancement processing by the image enhancement algorithm; Specifically, various algorithm data corresponding to various image enhancement algorithms are entered, and the single algorithm data corresponding to each image enhancement algorithm is the code length, type identification and processed reference image data after image enhancement processing by the image enhancement algorithm, including: the code of each image enhancement algorithm is a computer executable instruction for implementing the corresponding type of image enhancement algorithm; A model building device, used to perform multiple learning on the feedforward neural network to obtain a feedforward neural network after performing multiple learning and output it as an AI prediction model, wherein the number of learning times of the multiple learning is monotonically positively correlated with the pixel points of the reference image; For example, multiple learning is performed on the feedforward neural network to obtain the feedforward neural network after the multiple learning is performed and output as the AI ​​prediction model, and the monotonically positive correlation between the number of learning times of the multiple learning and the pixel points of the reference image includes: when the number of pixels of the reference image is 3840×2160, the value of the number of learning times of the multiple learning is 300, when the number of pixels of the reference image is 2560×1440, the value of the number of learning times of the multiple learning is 200, when the number of pixels of the reference image is 1920×1080, the value of the number of learning times of the multiple learning is 150, when the number of pixels of the reference image is 1600×900, the value of the number of learning times of the multiple learning is 120, when the number of pixels of the reference image is 1280×720, the value of the number of learning times of the multiple learning is 100, and so on; An effect prediction device is connected to the chip acquisition device, the content analysis device, the data entry device and the model building device respectively, and is used to use the AI ​​prediction model to intelligently predict the signal-to-noise ratio of the enhanced processed image obtained after the image to be enhanced is subjected to image enhancement processing by the target image enhancement algorithm based on multiple analysis parameters of the reference image, image data of the reference image, various acquisition information of the image processing chip, multiple analysis parameters of the image to be enhanced and algorithm data corresponding to the target image enhancement algorithm; an algorithm selection device, connected to the effect prediction device, for selecting the optimal image enhancement algorithm of the image to be enhanced based on the signal-to-noise ratios corresponding to the various image enhancement algorithms for execution by the image processing chip; Among them, selecting the optimal image enhancement algorithm of the image to be enhanced based on the signal-to-noise ratios corresponding to the various image enhancement algorithms for the image processing chip to execute includes: the various image enhancement processes corresponding to the various image enhancement algorithms are histogram correction, enhancement process based on linear grayscale transformation, enhancement process based on piecewise linear grayscale transformation, enhancement process based on logarithmic transformation, enhancement process based on exponential transformation and frequency domain enhancement process, and the optimal image enhancement algorithm of the image to be enhanced is the image enhancement algorithm corresponding to the signal-to-noise ratio with the largest value among the signal-to-noise ratios; Among them, performing multiple learning on the feedforward neural network to obtain the feedforward neural network after performing multiple learning and outputting it as the AI ​​prediction model, and the number of learning times of the multiple learning is monotonically positively correlated with the pixel points of the reference image, including: in each learning performed on the feedforward neural network, using the known signal-to-noise ratio of an enhanced processed image obtained after an image is enhanced by a certain image enhancement algorithm as the output content of the feedforward neural network, using multiple analytical parameters of the reference image, image data of the reference image, various acquisition information of the image processing chip, multiple analytical parameters of the certain image, and algorithm data corresponding to the certain image enhancement algorithm as multiple input contents of the feedforward neural network to perform this learning; Among them, the AI ​​prediction model is used to intelligently predict the signal-to-noise ratio of the enhanced processed image obtained after the image to be enhanced is subjected to image enhancement processing by the target image enhancement algorithm based on multiple analytical parameters of the reference image, image data of the reference image, various acquisition information of the image processing chip, multiple analytical parameters of the image to be enhanced, and algorithm data corresponding to the target image enhancement algorithm, including: the image data of the reference image is each pixel value corresponding to each pixel point of the reference image; For example, the image data of the reference image are pixel values ​​corresponding to the pixels of the reference image, including: the pixel values ​​corresponding to the pixels of the reference image are brightness values ​​corresponding to the pixels of the reference image; And wherein, the AI ​​prediction model is used to intelligently predict the signal-to-noise ratio of the enhanced processed image obtained after the image to be enhanced is processed by the target image enhancement algorithm based on multiple analytical parameters of the benchmark image, the image data of the benchmark image, various acquisition information of the image processing chip, multiple analytical parameters of the image to be enhanced, and the algorithm data corresponding to the target image enhancement algorithm, and also includes: the target image enhancement algorithm is histogram correction, enhancement processing based on linear grayscale transformation, enhancement processing based on piecewise linear grayscale transformation, enhancement processing based on logarithmic transformation, enhancement processing based on exponential transformation, and frequency domain enhancement processing. The algorithm corresponds to any one of the image enhancement processing.

[0031] Example 5 Figure 6 1 is an internal structure diagram of an image enhancement system based on an image processing chip according to Embodiment 5 of the present invention.

[0032] like Figure 6 As shown, the image enhancement system based on the image processing chip also includes: A data storage chip, connected to the model building device, is used to receive various model parameters of the AI ​​prediction model, and complete the model storage of the AI ​​prediction model by storing various model parameters of the AI ​​prediction model; By way of example, a data storage chip is connected to the model building device, and is used to receive various model parameters of the AI ​​prediction model, and complete the model storage of the AI ​​prediction model by storing the various model parameters of the AI ​​prediction model, including: the data storage chip is a TF storage chip, an MMC storage chip or a FLASH storage chip.

[0033] Example 6 Figure 7 1 is an internal structure diagram of an image enhancement system based on an image processing chip according to Embodiment 6 of the present invention.

[0034] like Figure 7 As shown, the image enhancement system based on the image processing chip also includes: An LCD display array, connected to the algorithm selection device, for receiving the optimal image enhancement algorithm of the image to be enhanced, and performing on-site display of the optimal image enhancement algorithm of the image to be enhanced; For example, an LCD display array is connected to the algorithm selection device, and is used to receive the optimal image enhancement algorithm for the image to be enhanced, and perform an on-site display of the optimal image enhancement algorithm for the image to be enhanced, including: a liquid crystal display screen or an LED display array can be selected to replace the LCD display array, and is used to perform an on-site display of the optimal image enhancement algorithm for the image to be enhanced.

[0035] And in any one of the above embodiments 4-6, optionally, in the image enhancement system based on the image processing chip: The signal-to-noise ratio, the number of high-amplitude noise types, the horizontal resolution, the vertical resolution, the pixel value redundancy and the number of main objects of the image to be enhanced are analyzed to output as multiple analytical parameters of the image to be enhanced, including: the high-amplitude noise type of the image to be enhanced is a noise type whose amplitude in the image to be enhanced exceeds a preset noise amplitude; Wherein, analyzing the signal-to-noise ratio, the number of high-amplitude noise types, the horizontal resolution, the vertical resolution, the pixel value redundancy and the number of main objects of the image to be enhanced as the multiple analytical parameter outputs of the image to be enhanced further includes: after the respective pixel values ​​corresponding to the respective pixel points of the image to be enhanced are deduplicated, the number of the remaining multiple pixel values ​​is a first number, the number of the respective pixel values ​​corresponding to the respective pixel points of the image to be enhanced is a second number, the first number is subtracted from the second number to obtain a third number, and the third number is divided by the second number to obtain the pixel value redundancy of the image to be enhanced; Wherein, analyzing the signal-to-noise ratio, the number of high-amplitude noise types, the horizontal resolution, the vertical resolution, the pixel value redundancy and the number of main objects of the image to be enhanced as the output of multiple analytical parameters of the image to be enhanced further includes: the main object of the image to be enhanced is an object in the image to be enhanced that occupies an area ratio of the image to be enhanced that is greater than or equal to a set ratio limit; The main object of the image to be enhanced is an object in the image to be enhanced that occupies an area ratio of the image to be enhanced that is greater than or equal to a set ratio limit, comprising: removing background pixels from pixels of the image to be enhanced to obtain foreground pixels, and performing fitting processing on the foreground pixels to obtain objects in the image to be enhanced; Specifically, the main object of the image to be enhanced is an object in the image to be enhanced that occupies an area ratio of the image to be enhanced greater than or equal to a set ratio limit, and further includes: the main object of the image to be enhanced may be a relatively large human body, tree branch or house, etc.; Among them, the algorithm data corresponding to various image enhancement algorithms are entered respectively, and the single algorithm data corresponding to each image enhancement algorithm is the code length, type identification and processed reference image data after image enhancement processing by the image enhancement algorithm, including: the code length of each image enhancement algorithm is the number of bits of binary data formed after the code of the image enhancement algorithm is converted into binary value; Among them, the algorithm data corresponding to various image enhancement algorithms are entered, and the single algorithm data corresponding to each image enhancement algorithm is the code length and type identification of the image enhancement algorithm, and the processed reference image data after the image enhancement processing of the image enhancement algorithm also includes: the processed reference image data after the image enhancement processing of each image enhancement algorithm is the pixel values ​​corresponding to each pixel point of the image obtained after the image enhancement processing of the image enhancement algorithm is performed on the reference image; Among them, performing multiple learning on the feedforward neural network to obtain the feedforward neural network after performing the multiple learning and outputting it as the AI ​​prediction model, wherein the number of learning times of the multiple learning and the monotonic positive correlation of the pixel points of the reference image further includes: using a numerical mapping function to represent the numerical mapping relationship of the monotonic positive correlation of the number of learning times of the multiple learning and the pixel points of the reference image; The method of using a numerical mapping function to represent the numerical mapping relationship of the monotonically positive correlation between the number of learning times of the multiple learning and the pixel points of the reference image comprises: in the numerical mapping function, the number of learning times of the multiple learning is the output parameter of the numerical mapping function, and the number of pixel points of the reference image is the input parameter of the numerical mapping function; Among them, using the AI ​​prediction model to intelligently predict the signal-to-noise ratio of the enhanced processed image obtained after the image to be enhanced is subjected to image enhancement processing by the target image enhancement algorithm based on multiple analytical parameters of the reference image, image data of the reference image, various acquisition information of the image processing chip, multiple analytical parameters of the image to be enhanced, and algorithm data corresponding to the target image enhancement algorithm also includes: synchronously inputting the multiple analytical parameters of the reference image, the image data of the reference image, various acquisition information of the image processing chip, multiple analytical parameters of the image to be enhanced, and algorithm data corresponding to the target image enhancement algorithm into the AI ​​prediction model; And wherein, using the AI ​​prediction model to intelligently predict the signal-to-noise ratio of the enhanced processed image obtained after the image to be enhanced is subjected to image enhancement processing by the target image enhancement algorithm based on multiple analytical parameters of the baseline image, the image data of the baseline image, various acquisition information of the image processing chip, multiple analytical parameters of the image to be enhanced, and algorithm data corresponding to the target image enhancement algorithm also includes: executing the AI ​​prediction model to obtain the signal-to-noise ratio of the enhanced processed image obtained after the image to be enhanced is subjected to image enhancement processing by the target image enhancement algorithm, which is output by the AI ​​prediction model.

[0036] In addition, in an image enhancement method and system based on an image processing chip according to the present invention: The AI ​​prediction model is used to intelligently predict the signal-to-noise ratio of the enhanced processed image obtained after the image to be enhanced is subjected to image enhancement processing by the target image enhancement algorithm based on multiple analytical parameters of the reference image, image data of the reference image, various acquisition information of the image processing chip, multiple analytical parameters of the image to be enhanced, and algorithm data corresponding to the target image enhancement algorithm, further comprising: performing numerical normalization processing on the multiple analytical parameters of the reference image, the image data of the reference image, various acquisition information of the image processing chip, multiple analytical parameters of the image to be enhanced, and algorithm data corresponding to the target image enhancement algorithm, respectively, and then synchronously inputting them into the AI ​​prediction model; For example, performing numerical normalization processing on multiple analytical parameters of the reference image, image data of the reference image, various acquisition information of the image processing chip, multiple analytical parameters of the image to be enhanced, and algorithm data corresponding to the target image enhancement algorithm and then synchronously inputting them into the AI ​​prediction model includes: the numerical normalization processing is a hexadecimal numerical conversion processing; Among them, in each learning performed on the feedforward neural network, the known signal-to-noise ratio of an enhanced processed image obtained after an image is enhanced by a certain image enhancement algorithm is used as the output content of the feedforward neural network, and multiple analytical parameters of the reference image, image data of the reference image, various acquisition information of the image processing chip, multiple analytical parameters of the certain image and algorithm data corresponding to the certain image enhancement algorithm are used as multiple input contents of the feedforward neural network. The execution of this learning also includes: choosing to use the MATLAB toolbox to realize the simulation and emulation of each learning performed on the feedforward neural network.

[0037] The above description is only a preferred embodiment of the present invention and is not intended to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention are included in the protection scope of the present invention.

Claims

1. An image enhancement method based on an image processing chip, running on the image processing chip, characterized in that: The method comprises: Collecting the memory capacity, operating frequency, fastest memory read and write speed, chip weight and chip volume of the image processing chip and outputting them as various collected information of the image processing chip; Analyze the signal-to-noise ratio, the number of high-amplitude noise types, the horizontal resolution, the vertical resolution, the pixel value redundancy, and the number of main objects of the image to be enhanced to output as multiple analytical parameters of the image to be enhanced; Enter the algorithm data corresponding to each image enhancement algorithm. The algorithm data corresponding to each image enhancement algorithm is the code length, type identification and processed reference image data after image enhancement processing by the image enhancement algorithm. Performing multiple learning on the feedforward neural network to obtain a feedforward neural network after performing the multiple learning and outputting it as an AI prediction model, wherein the number of learning times of the multiple learning is monotonically positively correlated with the pixel points of the reference image; The AI ​​prediction model is used to intelligently predict the signal-to-noise ratio of an enhanced processed image obtained after the image to be enhanced is subjected to image enhancement processing by the target image enhancement algorithm based on multiple analytical parameters of the reference image, image data of the reference image, various acquisition information of the image processing chip, multiple analytical parameters of the image to be enhanced, and algorithm data corresponding to the target image enhancement algorithm; Selecting the optimal image enhancement algorithm for the image to be enhanced based on the signal-to-noise ratios corresponding to the various image enhancement algorithms for execution by the image processing chip; Among them, selecting the optimal image enhancement algorithm of the image to be enhanced based on the signal-to-noise ratios corresponding to the various image enhancement algorithms for execution by the image processing chip includes: the various image enhancement processing corresponding to the various image enhancement algorithms are histogram correction, enhancement processing based on linear grayscale transformation, enhancement processing based on piecewise linear grayscale transformation, enhancement processing based on logarithmic transformation, enhancement processing based on exponential transformation and frequency domain enhancement processing, and the optimal image enhancement algorithm of the image to be enhanced is the image enhancement algorithm corresponding to the signal-to-noise ratio with the largest value among the various signal-to-noise ratios.

2. The image enhancement method based on an image processing chip according to claim 1, characterized in that: Performing multiple learning on the feedforward neural network to obtain the feedforward neural network after the multiple learning is performed and output as the AI ​​prediction model, wherein the number of learning times of the multiple learning is monotonically positively correlated with the pixel points of the reference image, including: in each learning performed on the feedforward neural network, using the known signal-to-noise ratio of an enhanced processed image obtained after an image is enhanced by a certain image enhancement algorithm as the output content of the feedforward neural network, using multiple analytical parameters of the reference image, image data of the reference image, various acquisition information of the image processing chip, multiple analytical parameters of the certain image, and algorithm data corresponding to the certain image enhancement algorithm as multiple input contents of the feedforward neural network to perform this learning; Among them, the AI ​​prediction model is used to intelligently predict the signal-to-noise ratio of the enhanced processed image obtained after the image to be enhanced is subjected to image enhancement processing by the target image enhancement algorithm based on multiple analytical parameters of the reference image, image data of the reference image, various acquisition information of the image processing chip, multiple analytical parameters of the image to be enhanced, and algorithm data corresponding to the target image enhancement algorithm, including: the image data of the reference image is each pixel value corresponding to each pixel point of the reference image; Among them, the AI ​​prediction model is used to intelligently predict the signal-to-noise ratio of the enhanced processed image obtained after the image to be enhanced is processed by the target image enhancement algorithm based on multiple analytical parameters of the benchmark image, the image data of the benchmark image, various acquisition information of the image processing chip, multiple analytical parameters of the image to be enhanced, and the algorithm data corresponding to the target image enhancement algorithm, and also includes: the target image enhancement algorithm is histogram correction, enhancement processing based on linear grayscale transformation, enhancement processing based on piecewise linear grayscale transformation, enhancement processing based on logarithmic transformation, enhancement processing based on exponential transformation, and frequency domain enhancement processing. The algorithm corresponding to any one of the image enhancement processing.

3. The image enhancement method based on an image processing chip as claimed in claim 2, characterized in that: After performing multiple learning on the feedforward neural network to obtain a feedforward neural network after performing the multiple learning and outputting it as an AI prediction model, and the number of learning times of the multiple learning is monotonically positively correlated with the pixel points of the reference image, the method further includes: A data storage chip is used to receive various model parameters of the AI ​​prediction model, and the model storage of the AI ​​prediction model is completed by storing the various model parameters of the AI ​​prediction model.

4. The image enhancement method based on an image processing chip as claimed in claim 2, characterized in that: After selecting the optimal image enhancement algorithm for the image to be enhanced based on the signal-to-noise ratios corresponding to the various image enhancement algorithms for execution by the image processing chip, the method further includes: An optimal image enhancement algorithm for the image to be enhanced is received, and the optimal image enhancement algorithm for the image to be enhanced is displayed on-site through an LCD display array.

5. The image enhancement method based on an image processing chip as claimed in claim 4, characterized in that: The signal-to-noise ratio, the number of high-amplitude noise types, the horizontal resolution, the vertical resolution, the pixel value redundancy and the number of main objects of the image to be enhanced are analyzed to output as multiple analytical parameters of the image to be enhanced, including: the high-amplitude noise type of the image to be enhanced is a noise type whose amplitude in the image to be enhanced exceeds a preset noise amplitude; Wherein, analyzing the signal-to-noise ratio, the number of high-amplitude noise types, the horizontal resolution, the vertical resolution, the pixel value redundancy and the number of main objects of the image to be enhanced as the multiple analytical parameter outputs of the image to be enhanced further includes: after the respective pixel values ​​corresponding to the respective pixel points of the image to be enhanced are deduplicated, the number of the remaining multiple pixel values ​​is a first number, the number of the respective pixel values ​​corresponding to the respective pixel points of the image to be enhanced is a second number, the first number is subtracted from the second number to obtain a third number, and the third number is divided by the second number to obtain the pixel value redundancy of the image to be enhanced; Wherein, analyzing the signal-to-noise ratio, the number of high-amplitude noise types, the horizontal resolution, the vertical resolution, the pixel value redundancy and the number of main objects of the image to be enhanced as the output of multiple analytical parameters of the image to be enhanced further includes: the main object of the image to be enhanced is an object in the image to be enhanced that occupies an area ratio of the image to be enhanced that is greater than or equal to a set ratio limit; The main object of the image to be enhanced is an object in the image to be enhanced that occupies an area ratio of the image to be enhanced that is greater than or equal to a set ratio limit, comprising: removing background pixels from pixels of the image to be enhanced to obtain foreground pixels, and performing fitting processing on the foreground pixels to obtain objects in the image to be enhanced; Among them, the algorithm data corresponding to various image enhancement algorithms are entered respectively, and the single algorithm data corresponding to each image enhancement algorithm is the code length, type identification and processed reference image data after image enhancement processing by the image enhancement algorithm, including: the code length of each image enhancement algorithm is the number of bits of binary data formed after the code of the image enhancement algorithm is converted into binary value; Among them, the algorithm data corresponding to various image enhancement algorithms are entered, and the single algorithm data corresponding to each image enhancement algorithm is the code length and type identification of the image enhancement algorithm, and the processed reference image data after the image enhancement processing of the image enhancement algorithm also includes: the processed reference image data after the image enhancement processing of each image enhancement algorithm is the pixel values ​​corresponding to each pixel point of the image obtained after the image enhancement processing of the image enhancement algorithm is performed on the reference image; Among them, performing multiple learning on the feedforward neural network to obtain the feedforward neural network after performing the multiple learning and outputting it as the AI ​​prediction model, wherein the number of learning times of the multiple learning and the monotonic positive correlation of the pixel points of the reference image further includes: using a numerical mapping function to represent the numerical mapping relationship of the monotonic positive correlation of the number of learning times of the multiple learning and the pixel points of the reference image; The method of using a numerical mapping function to represent the numerical mapping relationship of the monotonically positive correlation between the number of learning times of the multiple learning and the pixel points of the reference image comprises: in the numerical mapping function, the number of learning times of the multiple learning is the output parameter of the numerical mapping function, and the number of pixel points of the reference image is the input parameter of the numerical mapping function; Among them, using the AI ​​prediction model to intelligently predict the signal-to-noise ratio of the enhanced processed image obtained after the image to be enhanced is subjected to image enhancement processing by the target image enhancement algorithm based on multiple analytical parameters of the reference image, image data of the reference image, various acquisition information of the image processing chip, multiple analytical parameters of the image to be enhanced, and algorithm data corresponding to the target image enhancement algorithm also includes: synchronously inputting the multiple analytical parameters of the reference image, the image data of the reference image, various acquisition information of the image processing chip, multiple analytical parameters of the image to be enhanced, and algorithm data corresponding to the target image enhancement algorithm into the AI ​​prediction model; Among them, using the AI ​​prediction model to intelligently predict the signal-to-noise ratio of an enhanced processed image obtained after the image to be enhanced is subjected to image enhancement processing by the target image enhancement algorithm based on multiple analytical parameters of the baseline image, image data of the baseline image, various acquisition information of the image processing chip, multiple analytical parameters of the image to be enhanced, and algorithm data corresponding to the target image enhancement algorithm also includes: executing the AI ​​prediction model to obtain the signal-to-noise ratio of the enhanced processed image obtained after the image to be enhanced is subjected to image enhancement processing by the target image enhancement algorithm, which is output by the AI ​​prediction model.

6. An image enhancement system based on an image processing chip, running on the image processing chip, characterized in that: The system comprises: A chip acquisition device, used to acquire the memory capacity, operating frequency, fastest memory read and write speed, chip weight and chip volume of the image processing chip and output it as various acquisition information of the image processing chip; A content analysis device, used for analyzing the signal-to-noise ratio, the number of high-amplitude noise types, the horizontal resolution, the vertical resolution, the pixel value redundancy and the number of main objects of the image to be enhanced to output as multiple analysis parameters of the image to be enhanced; A data entry device, used to enter various algorithm data corresponding to various image enhancement algorithms, wherein the single algorithm data corresponding to each image enhancement algorithm is the code length and type identification of the image enhancement algorithm and the processed reference image data after image enhancement processing by the image enhancement algorithm; A model building device, used to perform multiple learning on the feedforward neural network to obtain a feedforward neural network after performing multiple learning and output it as an AI prediction model, wherein the number of learning times of the multiple learning is monotonically positively correlated with the pixel points of the reference image; An effect prediction device is connected to the chip acquisition device, the content analysis device, the data entry device and the model building device respectively, and is used to use the AI ​​prediction model to intelligently predict the signal-to-noise ratio of the enhanced processed image obtained after the image to be enhanced is subjected to image enhancement processing by the target image enhancement algorithm based on multiple analysis parameters of the reference image, image data of the reference image, various acquisition information of the image processing chip, multiple analysis parameters of the image to be enhanced and algorithm data corresponding to the target image enhancement algorithm; an algorithm selection device, connected to the effect prediction device, for selecting the optimal image enhancement algorithm of the image to be enhanced based on the signal-to-noise ratios corresponding to the various image enhancement algorithms for execution by the image processing chip; Among them, selecting the optimal image enhancement algorithm of the image to be enhanced based on the signal-to-noise ratios corresponding to the various image enhancement algorithms for execution by the image processing chip includes: the various image enhancement processing corresponding to the various image enhancement algorithms are histogram correction, enhancement processing based on linear grayscale transformation, enhancement processing based on piecewise linear grayscale transformation, enhancement processing based on logarithmic transformation, enhancement processing based on exponential transformation and frequency domain enhancement processing, and the optimal image enhancement algorithm of the image to be enhanced is the image enhancement algorithm corresponding to the signal-to-noise ratio with the largest value among the various signal-to-noise ratios.

7. The image enhancement system based on an image processing chip as claimed in claim 6, characterized in that: Performing multiple learning on the feedforward neural network to obtain the feedforward neural network after the multiple learning is performed and output as the AI ​​prediction model, wherein the number of learning times of the multiple learning is monotonically positively correlated with the pixel points of the reference image, including: in each learning performed on the feedforward neural network, using the known signal-to-noise ratio of an enhanced processed image obtained after an image is enhanced by a certain image enhancement algorithm as the output content of the feedforward neural network, using multiple analytical parameters of the reference image, image data of the reference image, various acquisition information of the image processing chip, multiple analytical parameters of the certain image, and algorithm data corresponding to the certain image enhancement algorithm as multiple input contents of the feedforward neural network to perform this learning; Among them, the AI ​​prediction model is used to intelligently predict the signal-to-noise ratio of the enhanced processed image obtained after the image to be enhanced is subjected to image enhancement processing by the target image enhancement algorithm based on multiple analytical parameters of the reference image, image data of the reference image, various acquisition information of the image processing chip, multiple analytical parameters of the image to be enhanced, and algorithm data corresponding to the target image enhancement algorithm, including: the image data of the reference image is each pixel value corresponding to each pixel point of the reference image; Among them, the AI ​​prediction model is used to intelligently predict the signal-to-noise ratio of the enhanced processed image obtained after the image to be enhanced is processed by the target image enhancement algorithm based on multiple analytical parameters of the benchmark image, the image data of the benchmark image, various acquisition information of the image processing chip, multiple analytical parameters of the image to be enhanced, and the algorithm data corresponding to the target image enhancement algorithm, and also includes: the target image enhancement algorithm is histogram correction, enhancement processing based on linear grayscale transformation, enhancement processing based on piecewise linear grayscale transformation, enhancement processing based on logarithmic transformation, enhancement processing based on exponential transformation, and frequency domain enhancement processing. The algorithm corresponding to any one of the image enhancement processing.

8. The image enhancement system based on the image processing chip as claimed in claim 7, characterized in that: The system further comprises: A data storage chip is connected to the model building device, and is used to receive various model parameters of the AI ​​prediction model, and complete the model storage of the AI ​​prediction model by storing the various model parameters of the AI ​​prediction model.

9. The image enhancement system based on an image processing chip as claimed in claim 7, characterized in that: The system further comprises: The LCD display array is connected to the algorithm selection device and is used to receive the optimal image enhancement algorithm of the image to be enhanced and perform on-site display of the optimal image enhancement algorithm of the image to be enhanced.

10. The image enhancement system based on the image processing chip according to claim 9, characterized in that: The signal-to-noise ratio, the number of high-amplitude noise types, the horizontal resolution, the vertical resolution, the pixel value redundancy and the number of main objects of the image to be enhanced are analyzed to output as multiple analytical parameters of the image to be enhanced, including: the high-amplitude noise type of the image to be enhanced is a noise type whose amplitude in the image to be enhanced exceeds a preset noise amplitude; Wherein, analyzing the signal-to-noise ratio, the number of high-amplitude noise types, the horizontal resolution, the vertical resolution, the pixel value redundancy and the number of main objects of the image to be enhanced as the multiple analytical parameter outputs of the image to be enhanced further includes: after the respective pixel values ​​corresponding to the respective pixel points of the image to be enhanced are deduplicated, the number of the remaining multiple pixel values ​​is a first number, the number of the respective pixel values ​​corresponding to the respective pixel points of the image to be enhanced is a second number, the first number is subtracted from the second number to obtain a third number, and the third number is divided by the second number to obtain the pixel value redundancy of the image to be enhanced; Wherein, analyzing the signal-to-noise ratio, the number of high-amplitude noise types, the horizontal resolution, the vertical resolution, the pixel value redundancy and the number of main objects of the image to be enhanced as the output of multiple analytical parameters of the image to be enhanced further includes: the main object of the image to be enhanced is an object in the image to be enhanced that occupies an area ratio of the image to be enhanced that is greater than or equal to a set ratio limit; The main object of the image to be enhanced is an object in the image to be enhanced that occupies an area ratio of the image to be enhanced that is greater than or equal to a set ratio limit, comprising: removing background pixels from pixels of the image to be enhanced to obtain foreground pixels, and performing fitting processing on the foreground pixels to obtain objects in the image to be enhanced; Among them, the algorithm data corresponding to various image enhancement algorithms are entered respectively, and the single algorithm data corresponding to each image enhancement algorithm is the code length, type identification and processed reference image data after image enhancement processing by the image enhancement algorithm, including: the code length of each image enhancement algorithm is the number of bits of binary data formed after the code of the image enhancement algorithm is converted into binary value; Among them, the algorithm data corresponding to various image enhancement algorithms are entered, and the single algorithm data corresponding to each image enhancement algorithm is the code length and type identification of the image enhancement algorithm, and the processed reference image data after the image enhancement processing of the image enhancement algorithm also includes: the processed reference image data after the image enhancement processing of each image enhancement algorithm is the pixel values ​​corresponding to each pixel point of the image obtained after the image enhancement processing of the image enhancement algorithm is performed on the reference image; Among them, performing multiple learning on the feedforward neural network to obtain the feedforward neural network after performing the multiple learning and outputting it as the AI ​​prediction model, wherein the number of learning times of the multiple learning and the monotonic positive correlation of the pixel points of the reference image further includes: using a numerical mapping function to represent the numerical mapping relationship of the monotonic positive correlation of the number of learning times of the multiple learning and the pixel points of the reference image; The method of using a numerical mapping function to represent the numerical mapping relationship of the monotonically positive correlation between the number of learning times of the multiple learning and the pixel points of the reference image comprises: in the numerical mapping function, the number of learning times of the multiple learning is the output parameter of the numerical mapping function, and the number of pixel points of the reference image is the input parameter of the numerical mapping function; Among them, using the AI ​​prediction model to intelligently predict the signal-to-noise ratio of the enhanced processed image obtained after the image to be enhanced is subjected to image enhancement processing by the target image enhancement algorithm based on multiple analytical parameters of the reference image, image data of the reference image, various acquisition information of the image processing chip, multiple analytical parameters of the image to be enhanced, and algorithm data corresponding to the target image enhancement algorithm also includes: synchronously inputting the multiple analytical parameters of the reference image, the image data of the reference image, various acquisition information of the image processing chip, multiple analytical parameters of the image to be enhanced, and algorithm data corresponding to the target image enhancement algorithm into the AI ​​prediction model; Among them, using the AI ​​prediction model to intelligently predict the signal-to-noise ratio of an enhanced processed image obtained after the image to be enhanced is subjected to image enhancement processing by the target image enhancement algorithm based on multiple analytical parameters of the baseline image, image data of the baseline image, various acquisition information of the image processing chip, multiple analytical parameters of the image to be enhanced, and algorithm data corresponding to the target image enhancement algorithm also includes: executing the AI ​​prediction model to obtain the signal-to-noise ratio of the enhanced processed image obtained after the image to be enhanced is subjected to image enhancement processing by the target image enhancement algorithm, which is output by the AI ​​prediction model.

Images