Intelligent image processing system based on generative image

Abstract

The invention discloses an intelligent image processing system based on a generative image, relates to the technical field of intelligent image processing, and solves the technical problem that in the prior art, an initial image meeting authenticity cannot be subjected to optimization processing detection. Specifically, a generation module analyzes and screens real-time generation conditions, determines an image main body through a descriptive text provided by a user, and obtains a generated primary image; the editing module edits and modifies the generated primary image and obtains an edited and modified image after completing editing and modification; the authenticity evaluation module performs authenticity evaluation on the edited and modified graph, and obtains a real adjusted graph after completing authenticity regulation and control; and the optimization module performs optimization detection processing on the real adjusted image, acquires image optimization parameters, obtains an optimization detection coefficient, deduces whether the current real adjusted image is optimized or not according to a coefficient comparison result, and obtains an optimized image after the optimization is completed or the optimization is not needed.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent image processing, and in particular to an intelligent image processing system based on generative images. Background Art

[0002] Generative images are images automatically created by computer programs and algorithms; Generator network: Generative image technology is usually based on the generator model in deep learning; the task of the generator is to convert random noise or some initial input into an image with certain visual features; Learning and training: Through a large amount of training data, the generator model gradually masters the characteristics and patterns of the image; during the training process, the model continuously adjusts its own parameters to improve the quality and authenticity of the generated images.

[0003] However, in the prior art, the generative image technology cannot perform targeted extraction of initial images according to user needs and cannot modify the initial images. At the same time, it is impossible to perform authenticity evaluation when the initial images are edited and modified, resulting in a decrease in image processing efficiency. In addition, it is impossible to optimize the processing and detection of initial images that meet the authenticity requirements, resulting in a decrease in the characteristics of the initial images themselves.

[0004] In view of the above technical defects, a solution is now proposed. Summary of the invention

[0005] The purpose of the present invention is to solve the above-mentioned problems and to provide an image intelligent processing system based on generative images.

[0006] The purpose of the present invention can be achieved through the following technical solutions:

[0007] An image intelligent processing system based on generative images, comprising an image intelligent processing platform, wherein the image intelligent processing platform is communicatively connected with a production end and a discrimination end;

[0008] The production end sends the generation signal to the generation module, which analyzes and screens the real-time generation conditions, determines the image subject through the descriptive text provided by the user, and obtains the generated initial image;

[0009] The editing module edits and modifies the generated initial image, collects the scale parameters and image parameters on the image, and infers whether the generated initial image needs to be edited and modified based on the parameter comparison, and obtains the edited image after the editing and modification is completed or no editing and modification is required;

[0010] The discriminant end cooperates with the image processing, and the authenticity assessment module evaluates the authenticity of the edited image, collects the light and shadow control information and the color control information, and infers whether the authenticity is qualified based on the information analysis, and obtains the real adjustment image after the authenticity control is completed or the authenticity is qualified;

[0011] The optimization module performs optimization detection processing on the real image adjustment, collects image optimization parameters and obtains optimization detection coefficients, and infers whether the current real image adjustment is optimized based on the coefficient comparison results, and obtains the optimized image after the optimization is completed or no optimization is required;

[0012] The structural evaluation module evaluates the internal structure of the optimized image, collects display structure information and three-dimensional structure information, and infers whether the optimized image needs proportional adjustment based on information processing. After the proportional adjustment is completed or no adjustment is required, the display image is generated and sent to the user end.

[0013] As a preferred embodiment of the present invention, the user's descriptive text is subjected to text analysis, the nouns appearing in the text provided by the user are collected, and whether the current noun is the image subject is inferred based on database analysis. If the noun in the provided text is unique, the current noun is used as the image subject; at the same time, when the noun is not unique, the historical usage scenarios of the corresponding nouns are compared in the database. If the historical usage scenarios of the corresponding non-unique nouns do not have the same scenario, the corresponding nouns are divided into multiple image subjects; if the historical usage scenarios of the corresponding non-unique nouns have the same scenario, the corresponding nouns are divided into a single image subject, and the adjectives corresponding to the nouns in the descriptive text are extracted and collected, and the current image subject and image embellishments are generated in a targeted manner based on the collected adjectives. After completion, it is set to generate an initial image and send it to the editing module.

[0014] As a preferred embodiment of the present invention, the scale parameters and image parameters on the image are respectively the numerical ratio of the ratio of the image subject and image embellishments in the initial image generated according to the position specifications and the average ratio of the same type of subjects in the database in the same usage scenario, and the proportion of the corresponding image parameters of the image subject and image embellishments in the initial image generated according to the position specifications and the image parameters of the same type of pictures in the database.

[0015] As a preferred embodiment of the present invention, if the scale parameter on the image is within the scale value ratio threshold range, and the image parameter on the image does not exceed the type quantity ratio threshold, it is inferred that the currently generated initial image detection is qualified; if the scale parameter on the image is not within the scale value ratio threshold range, or the image parameter on the image exceeds the type quantity ratio threshold, it is inferred that the currently generated initial image detection is unqualified.

[0016] As a preferred embodiment of the present invention, the light and shadow control information and the color control information are respectively the non-matching area between the shadow trajectory and the light irradiation trajectory of the image subject and the image embellishment under the light orientation given by the light source in the current edited image, and the maximum color difference of the color transition at the overlapping boundary between the image subject and the image embellishment in the edited image.

[0017] As a preferred embodiment of the present invention, if the light and shadow control information exceeds the non-fitting area threshold, or the color control information exceeds the transition maximum color difference threshold, it is inferred that the current editing and modification of the image requires authenticity control; if the light and shadow control information does not exceed the non-fitting area threshold, and the color control information does not exceed the transition maximum color difference threshold, it is inferred that the current editing and modification of the image does not require authenticity control.

[0018] As a preferred embodiment of the present invention, the image optimization parameters include the cumulative value of the numerical deviations between the display parameters corresponding to all local images in the real adjustment image and the set parameter thresholds, the reciprocating floating span value corresponding to the sharpness deviations of adjacent boundaries of any local image in the real adjustment image, and the numerical ratio corresponding to the image clarity floating amount of the local image during the continuous image processing of the real adjustment image and the compression amount of the local image file.

[0019] As a preferred implementation of the present invention, if the optimization detection coefficient of the real map adjustment exceeds the optimization detection coefficient threshold, it is inferred that the real map adjustment needs to be optimized; if the optimization detection coefficient of the real map adjustment does not exceed the optimization detection coefficient threshold, it is inferred that the real map adjustment does not need to be optimized.

[0020] As a preferred embodiment of the present invention, the display structure information and the three-dimensional structure information are respectively the shortest spacing deviation value of any point between the image main body and the image embellishment in the optimized image under the picture display perspective, and the ratio of the spatial display distance between the image main body and the image embellishment in the optimized image under the three-dimensional display angle and the normal distance of the usage scene and the proportional deviation value corresponding to the image ratio.

[0021] As a preferred embodiment of the present invention, if the displayed structure information exceeds the shortest spacing deviation value threshold, or the three-dimensional structure information exceeds the ratio deviation value threshold, it is inferred that the structure evaluation of the current optimized image is abnormal;

[0022] If the displayed structural information does not exceed the shortest spacing deviation value threshold, and the three-dimensional structural information does not exceed the ratio deviation value threshold, it is inferred that the structural evaluation of the current optimized image is normal.

[0023] Compared with the prior art, the present invention has the following beneficial effects:

[0024] 1. In the present invention, the conditional requirements for generating images are accurately analyzed and inferred through the generation conditions, so that the generated image can meet the generation requirements, ensuring that the output image is highly similar to the user's required image, avoiding the basic output image of the generated image deviating from the user's requirements, resulting in an increase in the intensity of image intelligent processing, and also causing a decrease in the efficiency of the generated image processing;

[0025] The generated initial image is edited and modified. By editing, detecting and modifying the generated initial image, the positions of the generated subject and embellishments in the image are made more reasonable, so as to facilitate further image processing and improvement of the generated initial image, so as to improve the accuracy and rationality of image generation.

[0026] 2. In the present invention, authenticity evaluation is used to infer whether the authenticity and reality of the currently generated image meet the requirements, so as to ensure the availability of the current version of the image after the editing and modification is completed, and the edited image can be adjusted in time when the authenticity is abnormal, thereby reducing the increase in the cost of subsequent image processing execution, thereby generating an image that meets the requirements more accurately.

[0027] 3. In the present invention, the actual image characteristic parameters of the real image are optimized and processed to ensure that they can meet the image display requirements, thereby ensuring that the generated image will not be affected by the image display, and the image processing previously performed can maximize the satisfaction of the generated image. At this time, the probability of cost waste of the optimization processing is low, and the real image can be more efficiently displayed after the optimization processing is completed, which also prepares for the final step of structural evaluation;

[0028] The structural evaluation module evaluates the internal structure of the optimized image to infer whether the structure of the main body of the current optimized image is normal, so as to avoid the low display quality of the generated image caused by unreasonable structure of the main body of the image in the optimized image. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In order to facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0030] Figure 1 It is the overall principle block diagram of the present invention;

[0031] Figure 2 This is a principle block diagram of the production end of the present invention;

[0032] Figure 3 It is a principle block diagram of the determination end of the present invention;

[0033] Figure 4 The present invention is a flowchart of the production end and the determination end. DETAILED DESCRIPTION

[0034] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0035] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0036] See also Figure 1 As shown, an image intelligent processing system based on generative images includes an image intelligent processing platform, the image intelligent processing platform is communicatively connected with a production end and a discrimination end, wherein the production end and the discrimination end cooperate with each other in the processing system to produce high-efficiency images;

[0037] See also Figure 2-4 As shown in the figure, the production end includes a generation module, an editing module, and an optimization module; the discrimination end includes a authenticity assessment module and a structure assessment module; in the image generation process, the production end and the discrimination end cooperate with each other to cover the corresponding modules to achieve the generated image quality that meets the actual needs;

[0038] After the image intelligent processing platform receives the generation instruction, the production end sends the generation signal to the generation module; after receiving the generation signal, the generation module analyzes and screens the real-time generation conditions, and accurately analyzes and infers the condition requirements for generating images through the generation conditions, so that the generated image can meet the generation requirements, ensuring that the output image is highly similar to the user's required image, avoiding the basic output image of the generated image deviating from the user's requirements, resulting in an increase in the intensity of image intelligent processing, and also causing a decrease in the efficiency of generated image processing;

[0039] Perform text analysis on the user's descriptive text, collect the nouns that appear in the text provided by the user, and infer whether the current noun is the main body of the image based on the database analysis. If the noun in the provided text is unique, the current noun is used as the main body of the image, such as mountain, tree, river, etc. If the noun in the provided text is not unique, analyze the current noun range, and use the nouns in a large range as the main body of the image, and the nouns in a small range as the embellishments of the image; for example, when a mountain and a tree appear at the same time, the mountain is used as the main body of the image, and the tree is used as the embellishment.

[0040] At the same time, when the noun is not unique, the historical usage scenarios of the corresponding noun are compared in the database. The database is a backup library for the first and final drafts of generative image generation. If the historical usage scenarios of the corresponding non-unique nouns do not have the same scene, the corresponding nouns are divided into multiple image subjects and generated separately. For example, if the mountain and the computer appear in the user's description text at the same time, separate images are generated. Here, only the preliminary image is generated based on the user's descriptive text. If the user needs two nouns to be in the same scene, a secondary description is performed. In addition, the database needs a certain number of times to infer whether different nouns are in the same usage scenario.

[0041] If the historical usage scenarios of the corresponding non-unique nouns exist in the same scene, the corresponding nouns are divided into individual image subjects, and the adjectives corresponding to the nouns in the descriptive text are extracted for collection. The current image subject and image embellishments are generated in a targeted manner based on the adjectives such as location, color, and shape in the collected adjectives. After completion, it is set to generate the initial image and send it to the editing module;

[0042] After receiving the generated initial image, the editing module edits and modifies the generated initial image. By editing, detecting and modifying the generated initial image, the positions of the generated subject and embellishments in the image are made more reasonable, so as to further process and improve the generated initial image, so as to improve the accuracy and rationality of image generation.

[0043] The numerical ratio of the ratio of the image subject and image embellishments in the generated initial image according to the position specification to the average ratio of the same type of subjects in the database in the same usage scenario is obtained. At the same time, the image parameters corresponding to the image subject and image embellishments in the generated initial image according to the position specification and the image parameters of the same type of pictures in the database are obtained. The image parameters are represented by color, brightness, contrast, saturation, etc. The deviation data type ratio is represented by the number of data types with deviations in the image parameters. The numerical ratio of the ratio of the image subject and image embellishments in the generated initial image according to the position specification to the average ratio of the same type of subjects in the database in the same usage scenario and the deviation data type ratio corresponding to the image parameters of the same type of pictures in the database are marked as the proportion parameters on the image and the image parameters on the image, and are compared with the proportion value ratio threshold range and the type quantity ratio threshold respectively:

[0044] If the ratio of the image subject and the image embellishment in the generated initial image generated according to the position specification and the average ratio of the same type of subjects in the database in the same usage scenario are within the ratio value ratio threshold range, and the number of deviation data types of the image parameters corresponding to the image subject and the image embellishment in the generated initial image generated according to the position specification and the image parameters of the same type of pictures in the database do not exceed the type number ratio threshold, it is inferred that the current generated initial image is qualified, and the current generated initial image is marked as edited;

[0045] If the numerical ratio of the ratio of the image subject and the image embellishment in the generated initial image generated according to the position specification to the average ratio of the same type of subjects in the database in the same usage scenario is not within the ratio numerical ratio threshold range, or the number of deviation data types of the image parameters corresponding to the image subject and the image embellishment in the generated initial image generated according to the position specification and the image parameters of the same type of pictures in the database exceed the type number ratio threshold, it is inferred that the current generated initial image detection is unqualified, and the current generated initial image is edited and modified, that is, the position specifications of the image subject and the image embellishment are modified, and the image parameters in the current picture are adjusted at the same time. After completion, the current generated initial image is marked as edited and modified;

[0046] After the editing and modification is completed, the discriminant end cooperates with the image processing, and the authenticity assessment module performs authenticity assessment on the edited image. Through the authenticity assessment, it is inferred whether the authenticity and reality of the currently generated image meet the requirements, so as to ensure the availability of the current version of the image after the editing and modification is completed, and to make timely adjustments to the edited image when the authenticity is abnormal, so as to reduce the cost increase of subsequent image processing execution, so as to more accurately generate an image that meets the requirements;

[0047] Obtain the non-matching area between the shadow trajectory and the light irradiation trajectory of the image subject and the image embellishment under the light direction given by the light source in the current edited image, where the non-matching area is expressed as the illogicality of the area of ​​the light and the corresponding shadow. For example, if the light is directed toward the subject from the front, and the shadow area should be mainly distributed on the back of the subject, it is illogical if the current shadow area is on the front. At the same time, obtain the maximum color difference of the color transition at the corresponding overlapping boundary of the image subject and the image embellishment in the edited image, where the color difference refers to the difference in hue, saturation, brightness, etc. between two or more colors during the color comparison process;

[0048] The non-matching area between the shadow trajectory and the light irradiation trajectory of the image subject and the image embellishment under the light direction given by the light source in the current edited image, and the maximum color difference of the color transition at the overlapping boundary between the image subject and the image embellishment in the edited image are marked as light and shadow control information and color control information, respectively, and compared with the non-matching area threshold and the transition maximum color difference threshold, respectively:

[0049] If the non-matching area between the shadow trajectory and the light irradiation trajectory of the image subject and the image embellishment in the light direction given by the light source in the current edited image exceeds the non-matching area threshold, or the maximum color difference of the color transition at the overlapping boundary of the image subject and the image embellishment in the edited image exceeds the transition maximum color difference threshold, it is inferred that the current edited image needs to be adjusted for authenticity, and the light and shadow trajectory of the edited image is adjusted or the transition color is smoothly controlled. After the optimization is completed, the edited image is marked as a real adjustment image;

[0050] If the non-matching area between the shadow trajectory and the light irradiation trajectory of the image subject and the image embellishment in the light direction given by the light source in the current edited image does not exceed the non-matching area threshold, and the maximum color difference of the color transition at the corresponding overlapping boundary of the image subject and the image embellishment in the edited image does not exceed the maximum color difference threshold of the transition, it is inferred that the current edited image does not need to be adjusted for authenticity, and the edited image is directly marked as a true adjustment image;

[0051] After obtaining the real adjustment image, the production end continues image processing. The optimization module optimizes and detects the real adjustment image. By optimizing and detecting the real adjustment image, it ensures that the actual image characteristic parameters of the real adjustment image can meet the image display requirements, thereby ensuring that the generated image will not be affected by the image display. The image processing performed previously can maximize the satisfaction of the generated image. At this time, the probability of cost waste of the optimization processing is low. After the real adjustment image is optimized, the generated image can be displayed more efficiently, which also prepares for the final step of structural evaluation.

[0052] The generated real adjustment map is divided into regions of the same area according to each pixel point, and the composed regions are divided into real adjustment maps. After the division is completed, the tissue area is marked as a local map, and the cumulative value of the numerical deviation between the display parameters and the set parameter threshold of all local maps in the real adjustment map is collected, where the display parameter is represented by the resolution and clarity of the real adjustment map; at the same time, the reciprocating floating span value corresponding to the sharpness deviation of adjacent boundaries of any local map in the real adjustment map is obtained, and the cumulative value of the numerical deviation between the display parameters and the set parameter threshold of all local maps in the real adjustment map and the reciprocating floating span value corresponding to the sharpness deviation of adjacent boundaries of any local map in the real adjustment map are marked as PLJ and RFK respectively; the corresponding numerical ratio of the image clarity floating amount of the corresponding local map in the continuous image processing process of the real adjustment map and the compression amount of the local map file is obtained, and the corresponding numerical ratio of the image clarity floating amount of the corresponding local map in the continuous image processing process of the real adjustment map and the compression amount of the local map file is marked as SZB;

[0053] The above collected information is uniformly marked as image optimization parameters, and substituted into the formula to obtain the optimized detection coefficient of the real image adjustment, where the formula is: Wherein, YH is the optimized detection coefficient, vkj1, vjk2 and vjk3 are preset proportional coefficients, e is a natural constant, and β is an error correction factor, which is 1.01;

[0054] Compare the optimized detection coefficient of the actual adjustment image with the optimized detection coefficient threshold:

[0055] If the optimization detection coefficient of the real adjustment image exceeds the optimization detection coefficient threshold, it is inferred that the real adjustment image needs to be optimized, and the local image in the image of the real adjustment image is optimized and controlled to ensure the clarity of each local position; after the optimization is completed, the real adjustment image is marked as optimized;

[0056] If the optimization detection coefficient of the real map adjustment does not exceed the optimization detection coefficient threshold, it is inferred that the real map adjustment does not need to be optimized, and the real map adjustment is directly marked as optimized;

[0057] After the optimized image is generated, the discriminant end performs coordinated image processing, and the structure evaluation module evaluates the internal structure of the optimized image to infer whether the structure of the main body of the current optimized image is normal, so as to avoid the unreasonable structure of the main body of the image in the optimized image, resulting in low display quality of the generated image, deviation from the generated image required by the user, and reduced execution efficiency of the generated image;

[0058] The shortest spacing deviation value of the distance between the main image and the image embellishment at any point in the optimized image under the picture display perspective is obtained, and at the same time, the ratio of the spatial display distance between the main image and the image embellishment at the optimized image under the stereoscopic display angle and the normal distance of the usage scene and the ratio deviation value corresponding to the image ratio are obtained, and the shortest spacing deviation value of the distance between the main image and the image embellishment at any point in the optimized image under the picture display perspective and the ratio of the spatial display distance between the main image and the image embellishment at the optimized image under the stereoscopic display angle and the normal distance of the usage scene and the ratio deviation value corresponding to the image ratio are marked as display structure information and stereoscopic structure information, and are compared with the shortest spacing deviation value threshold and the ratio deviation value threshold respectively: wherein the normal distance of the usage scene is represented as a distance value within the spacing range between the same type of subjects in the database;

[0059] If the shortest distance deviation value of any point between the image subject and the image embellishment in the optimized image under the image display perspective exceeds the shortest distance deviation value threshold, or the ratio of the spatial display distance between the image subject and the image embellishment in the optimized image under the stereoscopic display angle and the normal distance of the usage scene and the ratio deviation value of the two corresponding to the image ratio exceeds the ratio deviation value threshold, it is inferred that the structural evaluation of the current optimized image is abnormal, and local ratio adjustment is required. After the adjustment is completed, the optimized image is marked as a generated display image and sent to the user end;

[0060] If the shortest distance deviation value of any point between the image subject and the image embellishment in the optimized image under the picture display perspective does not exceed the shortest distance deviation value threshold, and the ratio of the spatial display distance between the image subject and the image embellishment in the optimized image under the stereoscopic display angle and the normal distance of the usage scenario and the ratio deviation value corresponding to the image ratio do not exceed the ratio deviation value threshold, then it is inferred that the structural evaluation of the current optimized image is normal, and the optimized image is directly marked as a generated display image and sent to the user end.

[0061] The above formulas are obtained by collecting a large amount of data and performing software simulation, and a formula close to the actual value is selected. The coefficients in the formula are set by technicians in this field according to actual conditions;

[0062] When the present invention is used, the production end sends the generation signal to the generation module, the generation module analyzes and screens the real-time generation conditions, determines the image subject through the descriptive text provided by the user, and obtains the generated initial image; the editing module edits and modifies the generated initial image, collects the scale parameters and the image parameters on the image, and infers whether the generated initial image needs to be edited and modified according to the parameter comparison, and obtains the edited image after the editing and modification is completed or no editing and modification is required; the discrimination end performs coordinated image processing, the authenticity evaluation module evaluates the authenticity of the edited image, collects light and shadow control information and color control information, infers whether the authenticity is qualified according to the information analysis, and obtains the real adjustment image after the authenticity control is completed or the authenticity is qualified;

[0063] The optimization module performs optimization detection processing on the real adjustment image, collects image optimization parameters and obtains optimization detection coefficients, and infers whether the current real adjustment image is optimized based on the coefficient comparison results, and obtains the optimized image after the optimization is completed or no optimization is required; the structure evaluation module performs intra-image structure evaluation on the optimized image, collects display structure information and three-dimensional structure information, and infers whether the optimized image needs proportional adjustment based on information processing, and generates a display image after the proportional adjustment is completed or no adjustment is required and sends it to the user end.

[0064] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. An intelligent image processing system based on generative images, characterized in that: It includes an image intelligent processing platform, which is connected to a production end and a judgment end in communication; The production end sends the generation signal to the generation module, which analyzes and screens the real-time generation conditions, determines the image subject through the descriptive text provided by the user, and obtains the generated initial image; The editing module edits and modifies the generated initial image, collects the scale parameters and image parameters on the image, and infers whether the generated initial image needs to be edited and modified based on the parameter comparison, and obtains the edited image after the editing and modification is completed or no editing and modification is required; The discriminant end cooperates with the image processing, and the authenticity assessment module evaluates the authenticity of the edited image, collects the light and shadow control information and the color control information, and infers whether the authenticity is qualified based on the information analysis, and obtains the real adjustment image after the authenticity control is completed or the authenticity is qualified; The optimization module performs optimization detection processing on the real image adjustment, collects image optimization parameters and obtains optimization detection coefficients, and infers whether the current real image adjustment is optimized based on the coefficient comparison results, and obtains the optimized image after the optimization is completed or no optimization is required; The structural evaluation module evaluates the internal structure of the optimized image, collects display structure information and three-dimensional structure information, and infers whether the optimized image needs proportional adjustment based on information processing. After the proportional adjustment is completed or no adjustment is required, the display image is generated and sent to the user end.

2. The image intelligent processing system based on generative image according to claim 1, characterized in that: The user's descriptive text is analyzed, and the nouns that appear in the text provided by the user are collected. Based on the database analysis, it is inferred whether the current noun is the image subject. If the noun in the provided text is unique, the current noun is used as the image subject. At the same time, when the noun is not unique, the historical usage scenarios of the corresponding nouns are compared in the database. If the historical usage scenarios of the corresponding non-unique nouns do not have the same scenario, the corresponding nouns are divided into multiple image subjects; if the historical usage scenarios of the corresponding non-unique nouns have the same scenario, the corresponding nouns are divided into a single image subject, and the adjectives corresponding to the nouns in the descriptive text are extracted and collected. According to the collected adjectives, the current image subject and image embellishments are generated in a targeted manner. After completion, it is set to generate the initial image and send it to the editing module.

3. The image intelligent processing system based on generative image according to claim 1, characterized in that: The scale parameters and image parameters on the figure are respectively the numerical ratio of the ratio of the image subject and image embellishments in the initial image generated according to the position specifications to the average ratio of the same type of subjects in the database in the same usage scenario, and the proportion of the corresponding image parameters of the image subject and image embellishments in the initial image generated according to the position specifications to the image parameters of the same type of pictures in the database.

4. The image intelligent processing system based on generative image according to claim 3, characterized in that: If the scale parameter on the image is within the scale value ratio threshold range, and the image parameter on the image does not exceed the type quantity ratio threshold, it is inferred that the current generated initial image detection is qualified; if the scale parameter on the image is not within the scale value ratio threshold range, or the image parameter on the image exceeds the type quantity ratio threshold, it is inferred that the current generated initial image detection is unqualified.

5. The image intelligent processing system based on generative image according to claim 1, characterized in that: The light and shadow control information and the color control information are respectively the non-matching area between the shadow trajectory and the light irradiation trajectory of the image subject and the image embellishment under the light direction given by the light source in the current edited image, and the maximum color difference of the color transition at the overlapping boundary between the image subject and the image embellishment in the edited image.

6. The image intelligent processing system based on generative image according to claim 5, characterized in that: If the light and shadow control information exceeds the non-fitting area threshold, or the color control information exceeds the transition maximum color difference threshold, it is inferred that the current edited image needs to be adjusted for authenticity; If the light and shadow control information does not exceed the non-fitting area threshold, and the color control information does not exceed the transition maximum color difference threshold, it is inferred that the current editing image does not require authenticity control.

7. The image intelligent processing system based on generative image according to claim 1, characterized in that: The image optimization parameters include the cumulative value of the numerical deviation between the display parameters of all local images in the real adjustment image and the set parameter threshold, the reciprocating floating span value corresponding to the sharpness deviation of adjacent boundaries of any local image in the real adjustment image, and the numerical ratio of the image clarity floating amount of the local image during the continuous image processing of the real adjustment image to the compression amount of the local image file.

8. The image intelligent processing system based on generative image according to claim 7, characterized in that: If the optimization detection coefficient of the actual map adjustment exceeds the optimization detection coefficient threshold, it is inferred that the actual map adjustment needs to be optimized; if the optimization detection coefficient of the actual map adjustment does not exceed the optimization detection coefficient threshold, it is inferred that the actual map adjustment does not need to be optimized.

9. The image intelligent processing system based on generative image according to claim 1, characterized in that: The display structure information and the three-dimensional structure information are respectively the shortest spacing deviation value of any point between the image subject and the image embellishment in the optimized image under the picture display perspective, and the ratio of the spatial display distance between the image subject and the image embellishment in the optimized image under the three-dimensional display angle and the normal distance of the usage scenario and the proportional deviation value corresponding to the image ratio.

10. The image intelligent processing system based on generative image according to claim 9, characterized in that: If the displayed structural information exceeds the shortest spacing deviation value threshold, or the three-dimensional structural information exceeds the ratio deviation value threshold, it is inferred that the structural evaluation of the current optimized image is abnormal; If the displayed structural information does not exceed the shortest spacing deviation value threshold, and the three-dimensional structural information does not exceed the ratio deviation value threshold, it is inferred that the structural evaluation of the current optimized image is normal.

Images