A method for generating advertising pictures based on AIGC

Abstract

The present invention relates to the field of image generation, and in particular to an AIGC-based advertising image generation method, comprising: periodically detecting the proportion of associated elements and the complexity coefficient of associated elements of target generated images in each target image combination to determine the combination state of each target image combination; determining an image analysis strategy according to the combination state; when extracting preferred generated elements by an element combination analysis method, determining the preferred generated elements according to a reference evaluation coefficient, and adjusting the combination coefficient of key analyzed elements; when extracting preferred generated elements by a preference analysis method, determining the preferred generated elements of the target image combination according to a reference evaluation coefficient and a reference usage frequency; under image generation conditions, determining the target combination elements according to a preferred index, and determining a target element set according to a reference combination coefficient. The present invention improves the validity of image synthesis results and image synthesis efficiency.

Description

Technical Field

[0001] The present invention relates to the field of image generation, and in particular to an advertising image generation method based on AIGC. Background Art

[0002] Generating advertising images based on AIGC technology can effectively save human resources, but the generated image results often still need to be manually modified, which affects the efficiency of image synthesis. There are rich image elements in the synthesized image, but the existing image generation methods often fail to take into account the impact of different elements and element combinations in the synthesized image on the evaluation results of the synthesized image. Therefore, how to analyze the elements and element combinations in combination with the existing synthesized images and their evaluation results to determine their actual impact on the effectiveness of the synthesized image is an urgent problem to be solved by technical personnel in this field.

[0003] Chinese patent application publication number CN118096916A discloses an image generation method, device, equipment and medium, the method comprising: obtaining target text description information, the target text description information carries relevant description information of the target business scenario, inputting the target text description information into an image generation model, and after being processed by the image generation model, outputting a target image showing target business elements, the image generation model is trained using a pair of image-text samples with a corresponding relationship, the image samples in the image-text sample pair show target business elements, and the text description samples in the image-text sample pair are generated based on the corresponding image samples. However, the above scheme has the following problems: it fails to iteratively optimize the elements and element combinations used based on the evaluation of each image synthesis result, so as to continuously adjust the image synthesis process, resulting in low effectiveness of subsequent image synthesis results, thereby affecting the image synthesis efficiency. Summary of the invention

[0004] To this end, the present invention provides an advertising image generation method based on AIGC, so as to overcome the problem in the prior art that the elements contained therein are not analyzed according to past image synthesis records to determine the influence of different elements and element combinations on the image generation results, resulting in low effectiveness of subsequent image synthesis results and further resulting in low image synthesis efficiency.

[0005] To achieve the above object, the present invention provides an advertising image generation method based on AIGC, comprising:

[0006] Periodically testing the proportion of associated elements and the complexity coefficient of associated elements of target generated images in each target image combination to determine the combination state of each target image combination;

[0007] Determining an image analysis strategy according to the combination state, including extracting preferred generation elements for the target image combination using an element combination analysis method and a preference analysis method;

[0008] When the element combination analysis method is used to extract the preferred generating elements, the preferred generating elements are determined according to the reference evaluation coefficients of each key analysis element, and whether to adjust the combination coefficients of the key analysis elements is determined according to the difference values ​​of the evaluation coefficients;

[0009] Under the condition that the element extraction is completed, whether to adjust the regional diffusion index in the element extraction process is determined according to the element overlap coefficient;

[0010] When the preferred generating elements are extracted by the preference analysis method, the preferred generating elements of the target image combination are determined according to the reference evaluation coefficient and the reference usage frequency;

[0011] Under the image generation condition, the target combination element is determined according to the preferred index, and the target element set recommended for image generation is determined according to the reference combination coefficient.

[0012] Further, the combination state of each target image combination is determined according to the proportion of associated elements of each target generated image and the complexity coefficient of the associated elements;

[0013] For a single target image combination,

[0014] If the reference associated element ratio is greater than the preset associated element ratio or the reference complexity coefficient is greater than the preset reference complexity coefficient, it is determined that the target image combination is in the first preset combination state;

[0015] If the reference associated element ratio is less than or equal to the preset associated element ratio and the reference complexity coefficient is less than or equal to the preset reference complexity coefficient, it is determined that the target image combination is in the second preset combination state;

[0016] The complexity coefficient of the associated elements of any target generated image is determined according to the number of associated elements in each associated element region and the element region overlap.

[0017] Further, determining an image analysis strategy corresponding to the target image combination according to the combination state of each target image combination;

[0018] If a target image combination is in a first preset combination state, extracting the preferred generation element of the target image combination by using an element combination analysis method;

[0019] If a target image combination is in the second preset combination state, a preference analysis method is used to extract the preferred generation elements for the target image combination.

[0020] Furthermore, when an element combination analysis method is used to optimally generate element extraction for a target image combination, the element combination analysis method includes:

[0021] Perform combination association analysis on the main keywords of the target image combination, which includes:

[0022] Determine the related keywords of each subject keyword according to the information association reference value, and set the associated information combination of the target image combination according to each subject keyword and its related keywords;

[0023] Each subject keyword in any associated information combination has at least one related keyword.

[0024] Further, for any association information combination, the association information matching coefficient of each association element region is determined according to the matching element proportion and the reference association degree, and the key element region corresponding to the association information combination is determined according to the association information matching coefficient;

[0025] The associated element region whose associated information matching coefficient is greater than the preset associated information matching coefficient is recorded as the key element region.

[0026] Further, if a target image combination completes the determination of the associated information combination, a preferred analysis is performed on the key element area corresponding to each associated information combination, a preferred generation element is determined according to the reference evaluation coefficient of each key analysis element, and whether to adjust the combination coefficient of the key analysis element is determined according to the difference value of the evaluation coefficient of each key analysis element;

[0027] For a single key analysis element, if the evaluation coefficient difference value is greater than the preset evaluation coefficient difference value, the combination coefficient of the key analysis element and each combination analysis element is reduced and adjusted according to the reference evaluation difference value;

[0028] The reduction value of the combination coefficient is positively correlated with the reference evaluation difference;

[0029] The key analysis element is a matching element whose element coincidence degree in the key element region is greater than a preset element coincidence degree.

[0030] Further, under the condition that the element extraction is completed, it is determined whether to adjust the regional diffusion index in the element extraction process according to the element overlap coefficient of the image generated by each target;

[0031] If the element overlap coefficient of a target generated image is greater than a preset element overlap coefficient, the regional diffusion index in the element extraction process is reduced and adjusted according to the element overlap coefficient;

[0032] The decrease value of the regional diffusion index is positively correlated with the element overlap coefficient;

[0033] The element extraction completion condition is that any target image combination in the first preset combination state completes the extraction of the preferred generated element.

[0034] Furthermore, when a preference analysis method is used to extract the optimal generation element for a target image combination, the preference analysis method includes:

[0035] Determining a preferred generation coefficient of the associated elements included in the target image combination according to the reference evaluation coefficient and the reference usage frequency;

[0036] Recording the associated element whose preferred generation coefficient is greater than the preset preferred generation coefficient as the preferred generation element of the target image combination;

[0037] The preferred generation coefficient is positively correlated with the reference evaluation coefficient and the reference usage frequency respectively.

[0038] Furthermore, under the image generation condition, the elements to be combined are determined based on the required text content, and the target combination element is determined according to the preference index, and the elements to be combined whose preference index is greater than the preset preference index are recorded as the target combination element;

[0039] The image generation condition is to obtain a new demand text content, and complete the main keyword extraction and combination association analysis of the demand text content.

[0040] Further, according to the target combination elements, image generation recommendation is performed for the corresponding required text content, and the set of target combination elements contained in a target generation image is recorded as a target element set;

[0041] The reference combination coefficient of any target element set is greater than a preset reference combination coefficient, and the reference combination coefficient is determined according to the combination coefficients between the target combination elements in the target element set.

[0042] Compared with the prior art, the beneficial effect of the present invention lies in that the technical scheme of the present invention determines the combination state according to the proportion of associated elements of the target generated image contained in the target image combination and the complexity coefficient of the associated elements, and determines the image analysis strategy of the target image combination according to the combination state, thereby ensuring that the analysis process for each target image combination conforms to the actual situation, ensuring the accuracy of the determined preferred generating elements and the adjustment results of the combination coefficients, thereby improving the effectiveness of subsequent image synthesis results and the image synthesis efficiency.

[0043] Furthermore, the present invention determines the combination state of the target image combination according to the proportion of reference associated elements and the complexity coefficient of associated elements, which is used to characterize the number and distribution of elements contained in each target generated image in the target image combination, and then determines a targeted image analysis strategy to avoid waste of resources caused by a unified analysis process, thereby improving the processing efficiency of the image analysis process.

[0044] Furthermore, in the present invention, for the target image combination in the first preset combination state, the target generated image is analyzed based on the associated information combination of the target image combination, and the associated information matching coefficient is determined according to the matching element proportion and the reference correlation degree to determine the key element area. The associated information matching coefficient is used to characterize the degree of correlation between the associated element area and the associated information combination. Only the key element area is subjected to preferential analysis, which improves the processing efficiency of the image analysis process while ensuring the validity of the determined preferred generated elements.

[0045] Furthermore, in the present invention, a preferred analysis is performed on the key element area, the preferred generating elements of the key element area are determined according to the reference evaluation coefficients, and whether to adjust the combination coefficients between the key analysis elements and their combined analysis elements is determined according to the evaluation coefficient difference values. In addition to single different elements, different element combinations will also affect the evaluation results of image synthesis. By determining the preferred generating elements and the combination coefficients according to the evaluation results of the existing image generation records, the subsequently generated image results are avoided from being affected by the element combinations. The present invention improves the effectiveness of the image synthesis results and the image synthesis efficiency.

[0046] Furthermore, the present invention determines whether to adjust the regional diffusion index in the element extraction process according to the element overlap coefficient of each target generated image. For situations where the element edge environment is more complex, the segmentation effect of the element image is guaranteed by adjusting the regional diffusion index, and a directly usable element image is provided for the subsequent image synthesis process. The present invention improves the image synthesis efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a schematic diagram of the method for generating an advertisement promotion image based on AIGC of the present invention;

[0048] Figure 2 It is a flow chart of determining the combination state of the target image combination according to the reference associated element ratio and the associated element complexity coefficient of the present invention;

[0049] Figure 3 A flow chart of the present invention for determining an image analysis strategy according to a combination state of each target image combination;

[0050] Figure 4 The present invention is a flow chart for determining whether to adjust the combination coefficient according to the evaluation coefficient difference value. DETAILED DESCRIPTION

[0051] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0052] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0053] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the drawings. This is merely for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0054] In addition, it should be noted that in the description of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0055] See also Figures 1 to 4 As shown, the present invention provides an advertising image generation method based on AIGC, comprising:

[0056] Periodically testing the proportion of associated elements and the complexity coefficient of associated elements of target generated images in each target image combination to determine the combination state of each target image combination;

[0057] Determining an image analysis strategy according to the combination state, including extracting preferred generation elements for the target image combination using an element combination analysis method and a preference analysis method;

[0058] When the element combination analysis method is used to extract the preferred generating elements, the preferred generating elements are determined according to the reference evaluation coefficients of each key analysis element, and whether to adjust the combination coefficients of the key analysis elements is determined according to the difference values ​​of the evaluation coefficients;

[0059] Under the condition that the element extraction is completed, whether to adjust the regional diffusion index in the element extraction process is determined according to the element overlap coefficient;

[0060] When the preferred generating elements are extracted by the preference analysis method, the preferred generating elements of the target image combination are determined according to the reference evaluation coefficient and the reference usage frequency;

[0061] Under the image generation condition, the target combination element is determined according to the preferred index, and the target element set recommended for image generation is determined according to the reference combination coefficient.

[0062] Among them, the present invention applies a cyclic image analysis cycle, and the duration of the image analysis cycle can be determined by the user. The higher the user's requirements for the effectiveness of the image synthesis result, the shorter the duration of the image analysis cycle. The duration of an image analysis cycle is provided, and the image analysis cycle is 15 days. At the end time of each image analysis cycle, the proportion of associated elements and the complexity coefficient of associated elements of the target generated image in each target image combination that completes synthesis and synthesis evaluation within the cycle are detected, and the subsequent image analysis process is completed;

[0063] In the present invention, a number of target generated images are generated for a user demand text. The number of generated target generated images can be set by the user according to actual needs. The higher the user's requirements for the effectiveness of the image synthesis process, the greater the number of target generated images generated for a user demand text. The user demand text is the text content of the user's requirements for the generated images. For a single user demand text, when the synthesis of the target generated images corresponding to the user demand text is completed, the set of the obtained target generated images is recorded as a target image combination. The user completes the synthesis effect evaluation for the obtained target generated images to obtain the result evaluation coefficient of each target generated image.

[0064] The material library used for synthesizing the target generated image in the present invention includes the acquired image elements and other image resources. The image elements are image areas corresponding to different objects, animals, plants and people. The user can set several attribute tags for each image element. An example of the attribute tags set for an image element is provided. If the recognition result of an image element is rose, its attribute tags are rose, flower and plant. There are several subject keywords in the present invention. The subject keywords are various nouns corresponding to different objects, animals, plants and people. Any subject keyword can correspond to one or more image elements, and the corresponding image element is recorded as its associated element. For a single subject keyword, its associated element has the same attribute tag as the subject keyword. The user can set the attribute tags included in the present invention according to the subject keywords of each user demand text in the historical synthesis record. The attribute tags included in the present invention include but are not limited to: animals, plants, natural landscapes, buildings, tools, lions, elephants, flowers, trees, leaves, mountains, rivers, lakes, temples, bridges and vehicles. How to obtain the recognition results of each image element in the target generated image and how to obtain the subject keywords in the user demand text are easy to understand for those skilled in the art and are not described in detail here.

[0065] Several image analysis records are applied in the present invention, and any image analysis record records the reference associated element ratio, reference complexity coefficient, information associated reference value, associated information matching coefficient, element coincidence degree, evaluation coefficient difference value, reference evaluation coefficient, element overlap coefficient and preferred generation coefficient in at least one image analysis process for a target image combination, and each image analysis record corresponds to a qualified mark, which records whether the user's requirements for the validity of the image synthesis result meet the user's needs.

[0066] Specifically, the combination state of each target image combination is determined according to the proportion of associated elements of each target generated image and the complexity coefficient of the associated elements;

[0067] For a single target image combination,

[0068] If the reference associated element ratio is greater than the preset associated element ratio or the reference complexity coefficient is greater than the preset reference complexity coefficient, it is determined that the target image combination is in the first preset combination state;

[0069] If the reference associated element ratio is less than or equal to the preset associated element ratio and the reference complexity coefficient is less than or equal to the preset reference complexity coefficient, it is determined that the target image combination is in the second preset combination state;

[0070] The complexity coefficient of the associated elements of any target generated image is determined according to the number of associated elements in each associated element region and the element region overlap.

[0071] Among them, for a single target image combination, the corresponding user demand text is analyzed to extract the main keywords, and the attribute labels of each image element in the target generated image contained in the target image combination are identified. The associated elements of each main keyword are determined according to the identified attribute labels of each image element, and the image is divided for each target generated image in the target image combination to obtain a number of element analysis areas with the same area, and the element analysis area where the associated element of any main keyword exists is recorded as the associated element area;

[0072] For a single target generated image, the proportion of associated elements = the number of associated element regions contained in the target generated image / the number of element analysis regions contained in the target generated image. The associated element complexity coefficient is the sum of the product of the associated reference value of the target generated image and the overlap reference value and the corresponding influence coefficient. The associated reference value is the average value of the number of associated elements in each associated element region contained. The overlap reference value is the average value of the element region overlap in each associated element region contained. The element region overlap of a single associated element region is the sum of the lengths of the intersecting edges of each element. If there is a common edge line between two associated elements, the edge line is recorded as an element intersection edge. A value of the influence coefficient corresponding to the associated reference value and the overlap reference value is provided. The value of the influence coefficient corresponding to the associated reference value is 0.5, and the value of the influence coefficient corresponding to the overlap reference value is 0.5.

[0073] For a single target image combination, the reference associated element ratio is the average value of the associated element ratios of each target generated image in the target image combination, and the reference complexity coefficient is the average value of the associated element complexity coefficients of each target generated image in the target image combination. The user can determine the values ​​of the preset associated element ratio and the preset reference complexity coefficient according to the actual working scenario. For example, the user can set them according to the image analysis record. The higher the user's requirements for the effectiveness of the image synthesis result, the larger the value of the preset associated element ratio and the larger the value of the preset reference complexity coefficient. A method for determining the value of the preset associated element ratio is provided, and the image analysis record for extracting the preferred generated element for the target image combination using the element combination analysis method is recorded as the first reference record, and the minimum value of the reference associated element ratio in the first reference record that meets the user's requirements for the effectiveness of the image synthesis result is recorded as the preset associated element ratio. A method for determining the value of the preset reference complexity coefficient is provided, and the minimum value of the reference complexity coefficient in the first reference record that meets the user's requirements for the effectiveness of the image synthesis result is recorded as the preset reference complexity coefficient.

[0074] Specifically, an image analysis strategy corresponding to the target image combination is determined according to the combination state of each target image combination;

[0075] If a target image combination is in a first preset combination state, extracting the preferred generation element of the target image combination by using an element combination analysis method;

[0076] If a target image combination is in the second preset combination state, a preference analysis method is used to extract the preferred generation elements for the target image combination.

[0077] Specifically, when an element combination analysis method is used to optimize and generate element extraction for a target image combination, the element combination analysis method includes:

[0078] Perform combination association analysis on the main keywords of the target image combination, which includes:

[0079] Determine the related keywords of each subject keyword according to the information association reference value, and set the associated information combination of the target image combination according to each subject keyword and its related keywords;

[0080] Each subject keyword in any associated information combination has at least one related keyword.

[0081] Among them, a combination association analysis is performed on any target image combination in a first preset combination state to divide and obtain a number of association information combinations, and the information association reference value between any two main keywords of the target image combination is the number of times that the two main keywords in the demand text content of the target image combination have an association relationship, and it is determined that there is an association relationship between two main keywords in the same sentence structure and having a grammatical relationship. The grammatical relationship in the present invention includes but is not limited to: subject-predicate relationship and verb-object relationship. How to identify the grammatical relationship between the main keywords is easy for technicians in this field to understand and will not be elaborated here;

[0082] For any subject keyword, other subject keywords whose information association reference values ​​are greater than the preset information association reference value are recorded as related keywords of the subject keyword. The value of the preset information association reference value can be determined by the user according to the actual working scenario. For example, the user can set it according to the image analysis record. The higher the user's requirement for the effectiveness of the image synthesis result, the larger the value of the preset information association reference value. A method for setting the preset information association reference value is provided, and the average value of the information association reference values ​​of the related keywords in the first reference record that meets the user's requirement for the effectiveness of the image synthesis result is recorded as the preset information association reference value.

[0083] Specifically, for any association information combination, the association information matching coefficient of each association element region is determined according to the matching element proportion and the reference association degree, and the key element region corresponding to the association information combination is determined according to the association information matching coefficient;

[0084] The associated element region whose associated information matching coefficient is greater than the preset associated information matching coefficient is recorded as the key element region.

[0085] Wherein, for any combination of associated information, the proportion of matching elements and the reference correlation degree of each associated element region are detected. For a single associated element region, the proportion of matching elements = the number of matching elements in the associated element region / the number of main keywords in the associated information combination, the matching element is the associated element of each main keyword in the associated information combination, the reference correlation degree = the number of associated elements of each related keyword of the key keyword in the associated element region / the number of related keywords of the key keyword in the associated element region, the key keyword is the main keyword with the largest number of related keywords in the associated information combination, and the associated information matching coefficient = ln (proportion of matching elements × reference correlation degree);

[0086] The value of the preset association information matching coefficient can be determined by the user according to the actual working scenario. For example, the user can set it according to the image analysis record. The higher the user's requirement for the effectiveness of the image synthesis result, the larger the value of the preset association information matching coefficient. A method for determining the value of the preset association information matching coefficient is provided, and the minimum value of the association information matching coefficient of each key element area in the first reference record that meets the user's requirement for the effectiveness of the image synthesis result is recorded as the preset association information matching coefficient.

[0087] Specifically, if a target image combination completes the determination of the associated information combination, a preferred analysis is performed on the key element area corresponding to each associated information combination, a preferred generation element is determined according to the reference evaluation coefficient of each key analysis element, and whether to adjust the combination coefficient of the key analysis element is determined according to the difference value of the evaluation coefficient of each key analysis element;

[0088] For a single key analysis element, if the evaluation coefficient difference value is greater than the preset evaluation coefficient difference value, the combination coefficient of the key analysis element and each combination analysis element is reduced and adjusted according to the reference evaluation difference value;

[0089] The reduction value of the combination coefficient is positively correlated with the reference evaluation difference;

[0090] The key analysis element is a matching element whose element coincidence degree in the key element region is greater than a preset element coincidence degree.

[0091] Wherein, if a target image combination completes the determination of the associated information combination, a preferred analysis is performed on the key element areas corresponding to each associated information combination. For a single matching element, the element coincidence degree = the number of target generated images containing the matching element in the target image combination / the number of target generated images in the target image combination. The value of the preset element coincidence degree can be determined by the user according to the actual working scenario. For example, the user can set it according to the image analysis record. The higher the user's requirements for the validity of the image synthesis result, the greater the value of the preset element coincidence degree. A method for determining the value of the preset element coincidence degree is provided. The average value of the element coincidence degrees of each key analysis element in the first reference record that meets the user's requirements for the validity of the image synthesis result is recorded as the preset element coincidence degree.

[0092] For a single key analysis element, the target generated image of the key analysis element in the target image combination is recorded as the key analysis image, and the evaluation coefficient difference value of the key analysis element is = , n is the number of key analysis images in the target image combination, is the result evaluation coefficient of the i-th key analysis image, is a reference evaluation coefficient, which is the average value of the result evaluation coefficients of each key analysis image. The value of the preset evaluation coefficient difference value can be determined by the user according to the actual working scenario. For example, the user can set it according to the image analysis record. The higher the user's requirement for the validity of the image synthesis result, the smaller the value of the preset evaluation coefficient difference value. A method for setting the preset evaluation coefficient difference value is provided. The image analysis record adjusted for the combination coefficient is recorded as an evaluation reference record, and the average value of the evaluation coefficient difference values ​​in the evaluation reference record that meets the user's requirement for the validity of the image synthesis result is recorded as the preset evaluation coefficient difference value.

[0093] The reference evaluation difference = the maximum value of the result evaluation coefficient in the key analysis image - the reference evaluation coefficient. The associated elements in the same key analysis area as the key analysis elements are recorded as combined analysis elements. When the analysis of each target image combination within an image analysis cycle is completed, the combined coefficient between each key analysis element and its combined analysis element is the accumulation of each reduction value during the image analysis process. The combined coefficients between the image elements are all initial combined coefficients before adjustment, and the value of the initial combined coefficient is 1.

[0094] The key analysis elements whose reference evaluation coefficients are greater than the preset reference evaluation coefficients are recorded as preferred generation elements. The value of the preset reference evaluation coefficient can be determined by the user according to the actual working scenario. For example, the user can set it according to the image analysis record. The higher the user's requirements for the effectiveness of the image synthesis result, the larger the value of the preset reference evaluation coefficient. A method for setting the value of the preset reference evaluation coefficient is provided, and the average value of each preferred generation element in the first reference record that meets the user's requirements for the effectiveness of the image synthesis result is recorded as the preset reference evaluation coefficient.

[0095] Specifically, under the condition that the element extraction is completed, it is determined whether to adjust the regional diffusion index in the element extraction process according to the element overlap coefficient of the image generated by each target;

[0096] If the element overlap coefficient of a target generated image is greater than a preset element overlap coefficient, the regional diffusion index in the element extraction process is reduced and adjusted according to the element overlap coefficient;

[0097] The decrease value of the regional diffusion index is positively correlated with the element overlap coefficient;

[0098] The element extraction completion condition is that any target image combination in the first preset combination state completes the extraction of the preferred generated element.

[0099] Among them, for a single target generated image, the element overlap coefficient is the sum of the lengths of the intersection edges of each preferred element in the target generated image. If there is a common edge line between two preferred generated elements in the target generated image, the edge line of the overlapping part is recorded as an intersection edge of a preferred element. The value of the preset element overlap coefficient can be determined by the user according to the actual working scenario. For example, the user can set it according to the image analysis record. The higher the user's requirements for the effectiveness of the image synthesis result, the smaller the value of the preset element overlap coefficient. A method for setting the value of the preset element overlap coefficient is provided. The minimum value of the element overlap coefficient of each target generated image adjusted for the regional diffusion index in the element extraction process in the first reference record that meets the user's requirements for the effectiveness of the image synthesis result is recorded as the preset element overlap coefficient;

[0100] When extracting the optimal generation elements for the target generated image, the Markov chain technology is used for image segmentation. The probability density value of each Markov chain obtained is detected. If the probability density value is greater than the regional diffusion index, it is determined to iteratively update the currently obtained Markov chain. If the probability density value is less than or equal to the regional diffusion index, it is determined to perform image segmentation according to the currently obtained Markov chain.

[0101] The regional diffusion index is used to determine whether to perform iterative update on the currently acquired Markov chain. The value of the regional diffusion index that has not been reduced and adjusted is recorded as the initial regional diffusion index. The value of the initial regional diffusion index can be determined by the user according to the actual working scenario. For example, the user can set it according to the image analysis record. The higher the user's requirements for the effectiveness of the image synthesis result, the smaller the value of the initial regional diffusion index. A method for determining the value of the initial regional diffusion index is provided. The average value of the probability density value of the Markov chain obtained during each image segmentation is recorded as the initial regional diffusion index. How to use the Markov chain technology for image segmentation, how to perform probability density value detection on the Markov chain, and how to iteratively update the Markov chain are contents that are easy to understand for technicians in this field and are not described in detail here.

[0102] Specifically, when a preference analysis method is used to extract the optimal generation element for a target image combination, the preference analysis method includes:

[0103] Determining a preferred generation coefficient of the associated elements included in the target image combination according to the reference evaluation coefficient and the reference usage frequency;

[0104] Recording the associated element whose preferred generation coefficient is greater than the preset preferred generation coefficient as the preferred generation element of the target image combination;

[0105] The preferred generation coefficient is positively correlated with the reference evaluation coefficient and the reference usage frequency respectively.

[0106] Among them, for a single associated element, the reference evaluation coefficient is the average value of the result evaluation coefficients of the target generated images containing the associated element, the reference usage frequency is the number of optimal generated images containing the associated element in the image analysis record, the optimal generated image is the target generated image with the largest result evaluation coefficient in each target image combination, the preferred generation coefficient = ln (reference evaluation coefficient × reference usage frequency), the value of the preset preferred generation coefficient can be determined by the user according to the actual working scenario, for example, the user can set it according to the image analysis record, the higher the user's requirement for the effectiveness of the image synthesis result, the larger the value of the preset preferred generation coefficient, and a method for determining the value of the preset preferred generation coefficient is provided, the image analysis record for extracting the preferred generation element for a target image combination using a preference analysis method is recorded as a second reference record, and the minimum value of the preferred generation coefficient of each preferred generation element in the second reference record that meets the user's requirement for the effectiveness of the image synthesis result is recorded as the preset preferred generation coefficient.

[0107] Specifically, under the image generation condition, the elements to be combined are determined based on the required text content, and the target combination elements are determined according to the preferred index, and the elements to be combined whose preferred index is greater than the preset preferred index are recorded as the target combination elements;

[0108] The image generation condition is to obtain a newly added demand text content, and complete the main keyword extraction and combination association analysis of the demand text content.

[0109] Among them, main keywords are extracted for the demand text content, and the associated elements corresponding to each main keyword in the material library used to synthesize the target generated image are recorded as elements to be combined. For a single element to be combined, the preferred index is the number of times the element to be combined is recorded as a preferred generated element in the most recent image analysis cycle. The value of the preset preferred index can be determined by the user according to the actual work scenario. For example, the user can set it according to the image analysis record. The higher the user's requirement for the effectiveness of the image synthesis result, the larger the value of the preset preferred index. A value of the preset preferred index is provided, and the value of the preset preferred index is 45% of the number of target image combinations in the most recent image analysis cycle.

[0110] Specifically, image generation recommendation is performed according to the target combination elements for the corresponding required text content, and a set of target combination elements contained in a target generation image is recorded as a target element set;

[0111] The reference combination coefficient of any target element set is greater than a preset reference combination coefficient, and the reference combination coefficient is determined according to the combination coefficients between the target combination elements in the target element set.

[0112] Among them, for a single target element set, the reference combination coefficient is the average value of the combination coefficients between the target combination elements in the target element set. The value of the preset reference combination coefficient can be determined by the user according to the actual working scenario. For example, the user can set it according to the image analysis record. The higher the user's requirement for the effectiveness of the image synthesis result, the higher the value of the preset reference combination coefficient. A method for determining the value of the preset reference combination coefficient is provided, and the minimum value of the reference combination coefficient of each optimal generated image is recorded as the preset reference combination coefficient. The determined target element sets are sent to the user to complete the image generation recommendation for the user demand text currently being processed.

[0113] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

[0114] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for generating an advertising image based on AIGC, characterized in that: include: Periodically testing the proportion of associated elements and the complexity coefficient of associated elements of target generated images in each target image combination to determine the combination state of each target image combination; Determining an image analysis strategy according to the combination state, including extracting preferred generation elements for the target image combination using an element combination analysis method and a preference analysis method; When the element combination analysis method is used to extract the preferred generating elements, the preferred generating elements are determined according to the reference evaluation coefficients of each key analysis element, and whether to adjust the combination coefficients of the key analysis elements is determined according to the difference values ​​of the evaluation coefficients; Under the condition that the element extraction is completed, whether to adjust the regional diffusion index in the element extraction process is determined according to the element overlap coefficient; When the preferred generating elements are extracted by the preference analysis method, the preferred generating elements of the target image combination are determined according to the reference evaluation coefficient and the reference usage frequency; Under the image generation condition, the target combination element is determined according to the preferred index, and the target element set recommended for image generation is determined according to the reference combination coefficient; The proportion of associated elements = the number of associated element areas contained in the target generated image / the number of element analysis areas contained in the target generated image. The associated element complexity coefficient is the sum of the product of the associated reference value and the overlap reference value of the target generated image and the corresponding influence coefficient. The associated reference value is the average value of the number of associated elements in each associated element area contained therein, and the overlap reference value is the average value of the overlap of element areas in each associated element area contained therein.

2. The method for generating an advertisement image based on AIGC according to claim 1, characterized in that: Determine the combination state of each target image combination according to the proportion of associated elements of each target generated image and the complexity coefficient of the associated elements; For a single target image combination, If the reference associated element ratio is greater than the preset associated element ratio or the reference complexity coefficient is greater than the preset reference complexity coefficient, it is determined that the target image combination is in the first preset combination state; If the reference associated element ratio is less than or equal to the preset associated element ratio and the reference complexity coefficient is less than or equal to the preset reference complexity coefficient, it is determined that the target image combination is in the second preset combination state; The complexity coefficient of the associated elements of the generated image of any target is determined according to the number of associated elements in each associated element region and the overlap of the element regions; For a single target image combination, the reference associated element ratio is the average ratio of associated element ratios of each target generated image in the target image combination, and the reference complexity coefficient is the average ratio of associated element complexity coefficients of each target generated image in the target image combination; The element area overlap of a single associated element area is the sum of the lengths of the intersecting edges of each element. If there is a common edge line between two associated elements, the edge line is recorded as an element intersection edge.

3. The method for generating an advertisement image based on AIGC according to claim 2, characterized in that: Determine the image analysis strategy corresponding to the target image combination according to the combination state of each target image combination; If a target image combination is in a first preset combination state, extracting the preferred generation element of the target image combination by using an element combination analysis method; If a target image combination is in the second preset combination state, a preference analysis method is used to extract the preferred generation elements for the target image combination.

4. The method for generating an advertisement image based on AIGC according to claim 3, characterized in that: When an element combination analysis method is used to optimize and generate element extraction for a target image combination, the element combination analysis method includes: Perform combination association analysis on the main keywords of the target image combination, which includes: Determine the related keywords of each subject keyword according to the information association reference value, and set the associated information combination of the target image combination according to each subject keyword and its related keywords; Each subject keyword in any associated information combination has at least one related keyword.

5. The method for generating an advertisement image based on AIGC according to claim 4, characterized in that: For any combination of associated information, determine the associated information matching coefficient of each associated element area according to the matching element proportion and the reference association degree, and determine the key element area corresponding to the associated information combination according to the associated information matching coefficient; Recording the associated element region whose associated information matching coefficient is greater than the preset associated information matching coefficient as the key element region; For a single associated element area, the reference association degree = the number of associated elements of each related keyword of the key keyword in the associated element area / the number of related keywords of the key keyword in the associated element area, the key keyword is the main keyword with the largest number of related keywords in the associated information combination, and the associated information matching coefficient = ln (matching element proportion × reference association degree).

6. The method for generating an advertisement image based on AIGC according to claim 5, characterized in that: If a target image combination completes the determination of the associated information combination, a preferred analysis is performed on the key element area corresponding to each associated information combination, a preferred generation element is determined according to the reference evaluation coefficient of each key analysis element, and whether to adjust the combination coefficient of the key analysis element is determined according to the difference value of the evaluation coefficient of each key analysis element; For a single key analysis element, if the evaluation coefficient difference value is greater than the preset evaluation coefficient difference value, the combination coefficient of the key analysis element and each combination analysis element is reduced and adjusted according to the reference evaluation difference value; The reduction value of the combination coefficient is positively correlated with the reference evaluation difference; The key analysis element is a matching element whose element coincidence degree in the key element area is greater than a preset element coincidence degree; For a single matching element, the element overlap degree = the number of target generated images containing the matching element in the target image combination / the number of target generated images in the target image combination.

7. The method for generating an advertisement image based on AIGC according to claim 6, characterized in that: Under the condition that the element extraction is completed, whether to adjust the regional diffusion index in the element extraction process is determined according to the element overlap coefficient of the image generated by each target; If the element overlap coefficient of a target generated image is greater than a preset element overlap coefficient, the regional diffusion index in the element extraction process is reduced and adjusted according to the element overlap coefficient; The decrease value of the regional diffusion index is positively correlated with the element overlap coefficient; The element extraction completion condition is that any target image combination in the first preset combination state completes the extraction of the preferred generated element; For a single target generated image, the element overlap coefficient is the sum of the lengths of the intersection edges of each preferred element in the target generated image. If there is a common edge line between two preferred generated elements in the target generated image, the edge line of the overlapping part is recorded as a preferred element intersection edge. The regional diffusion index is used to determine whether to perform iterative update on the currently acquired Markov chain. The value of the regional diffusion index that has not been reduced is recorded as the initial regional diffusion index, and the average value of the probability density value of the Markov chain obtained during each image segmentation is recorded as the initial regional diffusion index.

8. The method for generating an advertisement image based on AIGC according to claim 4, characterized in that: When a preference analysis method is used to extract the optimal generation element for a target image combination, the preference analysis method includes: Determining a preferred generation coefficient of the associated elements included in the target image combination according to the reference evaluation coefficient and the reference usage frequency; Recording the associated element whose preferred generation coefficient is greater than the preset preferred generation coefficient as the preferred generation element of the target image combination; The preferred generation coefficient is positively correlated with the reference evaluation coefficient and the reference usage frequency respectively.

9. The method for generating an advertisement image based on AIGC according to claim 8, characterized in that: Under the image generation condition, the elements to be combined are determined based on the required text content, and the target combination elements are determined according to the preferred index, and the elements to be combined whose preferred index is greater than the preset preferred index are recorded as the target combination elements; The image generation condition is to obtain a new demand text content, and complete the main keyword extraction and combination association analysis of the demand text content.

10. The method for generating an advertisement image based on AIGC according to claim 9, characterized in that: An image generation recommendation is performed according to the target combination elements for the corresponding demand text content, and a set of target combination elements contained in a target generation image is recorded as a target element set; The reference combination coefficient of any target element set is greater than a preset reference combination coefficient, and the reference combination coefficient is determined according to the combination coefficients between the target combination elements in the target element set.

Images