Image evaluation system and image evaluation method

Abstract

It is possible to evaluate images that have been partially generated automatically. [Solution] The system is characterized by comprising: an information acquisition unit that acquires a target image obtained by applying an image generation process to a portion of an existing image; boundary information indicating the boundary between the generated region, which is the portion of the image generation process that was applied to the image generation process, and the existing image region that was not applied to the image generation process; a calculation unit that calculates the change in features at the boundary; and a determination unit that determines the state of the generated region from the change in features at the boundary.

Description

Technical Field

[0001] The present invention relates to an image evaluation system and an image evaluation method.

Background Art

[0002] Conventionally, there is a technique for evaluating an image in which a part is automatically generated. For example, in Japanese Patent Application Laid-Open No. 2016-206777 (Patent Document 1), "in an image obtained by performing a completion process on a missing area, reducing the visual discomfort inside and outside the missing area", "from a reference image stored in a reference image storage unit that stores a reference image used when searching for a similar area similar to the area including the missing area, a similar area similar to the area including the missing area is selected for each feature amount set based on a plurality of feature amount sets, and by using the similar area, a first image is generated for each feature amount set by performing a completion process on the missing area in the original image, and a ratio of the amount of gaze collected in the area corresponding to the missing area in the original image and the first image is calculated for each feature amount set, and an evaluation value that is the evaluation result of the naturalness of the first image is calculated for each feature amount set using the ratio of the amount of gaze, and a selection unit that selects the first image with the least visual discomfort inside and outside the missing area from the first images for each feature amount set".

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] In the conventional technique, it is evaluated how much an image obtained by performing a completion process on a missing area has changed compared to the original image. In the conventional technique, it does not consider whether the boundary between the automatically generated area and the area of the original image is smoothly connected without discomfort.

[0005] Therefore, the present invention aims to evaluate an image that has been partially automatically generated by focusing on the boundaries of the automatically generated region. [Means for solving the problem]

[0006] To achieve the above objective, one representative image evaluation system of the present invention is characterized by comprising: an information acquisition unit that acquires a target image obtained by applying an image generation process to a partial region of an existing image; boundary information indicating the boundary between the generated region, which is the partial region subjected to the image generation process, and the existing image region that is not subject to the image generation process; a calculation unit that calculates changes in features at the boundary; and a determination unit that determines the state of the generated region from the changes in features at the boundary. Furthermore, one representative image evaluation method of the present invention is characterized by including an information acquisition step in which a computer acquires a target image obtained by applying an image generation process to a partial region of an existing image, and boundary information indicating the boundary between the generated region, which is the partial region subjected to the image generation process, and the existing image region that is not subject to the image generation process; a calculation step in which changes in features at the boundary are calculated; and a determination step in which the state of the generated region is determined from the changes in features at the boundary. [Effects of the Invention]

[0007] According to the present invention, it is possible to evaluate images that have been partially automatically generated. Other problems, configurations, and effects will be clarified by the following description of embodiments. [Brief explanation of the drawing]

[0008] [Figure 1] Configuration diagram of the image evaluation system [Figure 2] Functional block diagram of the image evaluation system [Figure 3] Specific examples of data [Figure 4] Flowchart showing the processing procedure of the information acquisition unit [Figure 5]Flowchart showing the processing procedure of the calculation unit [Figure 6] A flowchart showing the procedure by which the scoring unit determines the quality of each edge. [Figure 7] A flowchart showing the procedure by which the score determination unit determines the ratio of low-scoring parts on each side. [Figure 8] Flowchart showing the processing procedure of the parameter search unit [Figure 9] A concrete example of displaying a list of evaluation results based on the evaluation of each side. [Figure 10] Specific examples of detailed image display based on the evaluation of each side. [Figure 11] A specific example of displaying a list of evaluation results based on the percentage of low-rated portions. [Figure 12] Specific examples of image detail display based on the percentage of negative feedback. [Figure 13] Specific example of an input reception screen by the search area setting unit [Figure 14] A concrete example of a list display of search results by the search result presentation unit. [Figure 15] Example of displaying detailed parameter set information [Modes for carrying out the invention]

[0009] The following describes an example using drawings. [Examples]

[0010] Figure 1 is a diagram showing the configuration of the image evaluation system. The image evaluation system 10 is a computer having a CPU (Central Processing Unit) 11, a main memory 12, an auxiliary storage device 13, and a network interface 14.

[0011] The image evaluation system 10 is connected to the user terminal 20 via a network. The user terminal 20 is a computer having a CPU 21, a memory 22, an auxiliary storage device 23, a network interface 24, an input device 25, and an output device 26. The input device 25 is, for example, a keyboard or a pointing device. The output device 26 is, for example, a liquid crystal display.

[0012] FIG. 2 is a functional block diagram of the image evaluation system. The image evaluation system 10 includes an image evaluation unit 30, an evaluation result presentation unit 40, and a parameter search unit 50. The image evaluation unit 30 includes an information acquisition unit 31, a calculation unit 32, and a score determination unit 33. The parameter search unit 50 includes a search area setting unit 51, a search unit 52, and a search result presentation unit 53. The functions of these processing units are realized by a processor (CPU 11 being an example) executing a program developed in the memory 12. Another example of the processor is a GPU (Graphics Processing Unit). Note that a part or all of the processing performed by the processor may be performed by an FPGA (Programmable Gate Array), or may be performed by other semiconductor integrated circuits.

[0013] The information acquisition unit 31 acquires an image and boundary information. The image acquired by the information acquisition unit 31 is an image obtained by performing an image generation process on a partial region of an existing image and is the object of evaluation. For the sake of convenience, the image that is the object of evaluation by the image evaluation unit 30 is referred to as the target image. The boundary information acquired by the information acquisition unit 31 indicates the boundary between the generation region, which is the partial region that is the object of the image generation process in the target image, and the existing image region that is not the object of the image generation process. The information acquisition unit 31 uses the target image and the boundary information and outputs the color information of the boundary to the calculation unit 32 as a list.

[0014] The calculation unit 32 calculates the change in features at the boundary using the boundary color information list. Specifically, the calculation unit 32 calculates the magnitude of the color change at the boundary. For example, the amount of change in the color vectors inside and outside the boundary may be calculated.

[0015] The score determination unit 33 is a determination unit that determines the state of the generated region from changes in features at the boundary. Based on the boundary color information list and boundary information, the score determination unit 33 detects portions where color changes are continuous along the boundary. For this detection, the score determination unit 33 uses a first threshold and a second threshold. The first threshold is a threshold for the amount of change in the color vector. The first threshold is called the change amount threshold. The second threshold is a threshold for the continuity along the boundary where the amount of change in the color vector exceeds the first threshold. The second threshold is called the continuity threshold. The score determination unit 33 gives a low evaluation to portions of the boundary where the amount of change in the color vector exceeds the first threshold (change amount threshold) and continues for more than the second threshold (continuity threshold). The portions that receive a low evaluation are portions that the user perceives as unnatural when viewed. The target images that the image evaluation system 10 gives a high evaluation to are images that are partially automatically generated and where the boundary between the generated region and the existing image region blends naturally. If there are any inconsistencies in the boundary between the generated area and the existing image area, the evaluation of that image will be lower.

[0016] The score determination unit 33 outputs boundary information and a score. The boundary information is the coordinate information of the portion (low-evaluation portion) where the amount of change in the color vector exceeds the first threshold (change threshold) and continues for a second threshold (continuity threshold) or more. The score is the evaluation result of the boundary.

[0017] Furthermore, the score determination unit 33 can evaluate the target image itself using a third threshold. For example, if the image judgment criterion, which is the third threshold, is set to "3", the target image itself will be given a low rating if there are 3 or more low-rated edges, and a high rating if there are fewer than 3 low-rated edges. A low-rated target image may be labeled "bad" as the judgment result, and a high-rated target image may be labeled "good" as the judgment result.

[0018] The evaluation result presentation unit 40 presents the user with boundary information and a score, indicating the evaluation result of the generated area. Specifically, the evaluation result presentation unit 40 displays an image on the user terminal 20 in which the boundary and the low-evaluated portion are superimposed on the target image. Alternatively, it may display the judgment result of the target image.

[0019] The search area setting unit 51 specifies the parameter range that the search unit 52 will search. The search unit 52 numerically searches for the optimal values ​​of the parameters set by the image evaluation unit 30. The parameters set by the image evaluation unit 30 include parameters set by at least one of the information acquisition unit 31, calculation unit 32, and score determination unit 33. The search unit 52 repeatedly performs image evaluation using the parameter range specified by the search area setting unit 51 and the evaluated image data to search for the optimal set of parameters. The evaluated image data is image data to which evaluation results such as good or bad have been pre-assigned as labels.

[0020] The search unit 52 reflects the search results in the image evaluation unit 30. The search unit 52 also outputs the search results to the search result presentation unit 53. The search result presentation unit 53 presents the search results from the search unit 52 to the user. This presentation to the user can be done, for example, by displaying it on the user terminal 20.

[0021] Figure 3 shows a specific example of the data. The boundary information indicates the boundary of the generation region. Here, the shape of the generation region is defined as a rectangle, and the boundary information is defined by setting the vertical Y coordinate values ​​and the horizontal X coordinate values ​​for the image name that identifies the image.

[0022] The boundary color information list is a list of color information for the inside and outside of the boundary. It stores RGB values ​​for pixels inside and outside each side of the generation region. Here, the focus is on pixels a predetermined number of pixels inside the boundary and pixels immediately outside the boundary. The predetermined number of pixels is a parameter set in the information acquisition unit 31. In the example in Figure 3, the top inner pixel on the left side of the image named "1_0_b_0.png" has RGB[124,223,56], and the outer pixel has RGB[125,226,59]. Similarly, the second inner pixel from the top on the left side of the image named "1_0_b_0.png" has RGB[124,224,56], and the outer pixel has RGB[126,227,64].

[0023] Boundary information indicates the portion of the boundary that receives a low evaluation. In the example in Figure 3, the left edge of the image named "1_0_b_0.png" is associated with [[610,640],[650,700]]. This indicates that, for the left edge of the image, the Y coordinates of 610-640 and 650-700 are regions that continuously exceed the threshold.

[0024] The score could be determined by judging the quality of each side of a rectangle, for example. In this case, each side would receive a score of either "bad" or "good". Alternatively, the score could be determined by judging the percentage of the low-rated portion of each side of the rectangle. In this case, each side would receive a score between 0.0 and 1.0.

[0025] Figure 4 is a flowchart showing the processing procedure of the information acquisition unit. The information acquisition unit 31 sequentially executes the following steps S101 to S104. In step S101, the information acquisition unit 31 sets parameters related to the distance between the points to be compared. Then, the process proceeds to step S102.

[0026] In step S102, the information acquisition unit 31 obtains the coordinates of the points adjacent to each edge of the generated region on the outside, and the coordinates of the points in the generated region X pixels inward from those points, from the boundary information. Then, the process proceeds to step S103. In step S103, the information acquisition unit 31 reads the color information of the corresponding coordinates inside and outside the boundary for each edge of the target image. Then, the process proceeds to step S104. In step S104, the information acquisition unit 31 transmits the color information of the inside and outside of the boundaries of each side to the calculation unit and terminates the process.

[0027] The parameter for determining the points to compare inside and outside the boundary should, for example, be set to specify two points beyond the blurred range when using an image generation process that blurs the boundary of the generation region. In Figure 4, the generation region is represented as a rectangle, and "X" is used as the parameter for explanation.

[0028] If the generation area is rectangular, the outer edge is fixed one pixel outside the generation area, and points a few pixels inside the generation area are used as the comparison point. If the generation area is not limited to a rectangle and a general generation area is assumed, a shape identical in shape to the generation area but with an area Y% smaller on the inside can be prepared, and the point closest to the outer point can be used for comparison. In this case, Y is set as a parameter. The color information is, for example, an RGB color vector, but other color systems may also be used.

[0029] Figure 5 is a flowchart showing the processing procedure of the calculation unit. The calculation unit 32 sequentially executes the following steps S201 to S205. In step S201, the calculation unit 32 obtains the color vectors inside and outside the boundaries of each edge from the information acquisition unit 31. Then, the process proceeds to step S202. Step S202: The calculation unit 32 sets the color space. Then, the process proceeds to step S203.

[0030] Steps S203 to S204 are loop operations for each edge. Step S203: The calculation unit 32 calculates a difference vector for the color vectors inside and outside the boundary. Then, the process proceeds to step S204. Step S204: The calculation unit 32 calculates the absolute value of all patterns for each element of the difference vector of the color vectors inside and outside the boundary, and then calculates their sum. That is, it finds the absolute value of the difference vector between R and G, the absolute value of the difference vector between G and B, and the absolute value of the difference vector between B and R, and then sums them up. After finding this sum for each pixel on each edge, the process proceeds to step S205. Step S205: The calculation unit 32 sends a list of sums for each side to the score determination unit 33 and terminates the process.

[0031] Figure 6 is a flowchart showing the procedure by which the score determination unit determines the quality of each edge. The score determination unit 33 sequentially executes the following steps S301 to S307. Step S301: The score determination unit 33 sets the change threshold and the continuity threshold. Then, the process proceeds to step S302. Step S302: The score determination unit 33 obtains boundary information and a list of sums for each edge. Then, the process proceeds to step S303.

[0032] Steps S303 to S306 are loop operations for each edge. In step S303, the score determination unit 33 determines whether the sum value exceeds the change threshold, and if it exceeds it continuously, it calculates the length and coordinate information (boundary information). After that, the process proceeds to step S304. Step S304 The score determination unit 33 determines whether the length of time that continuously exceeds the change threshold exceeds the continuity threshold. If it does not, proceed to step S305. If it does, proceed to step S306. Step S305: The score determination unit 33 determines the edge in question as "good". If processing is complete for each edge, proceed to step S307. If there are any unprocessed edges, return to step S303 and loop. Step S306: The score determination unit 33 determines the edge in question as "bad". If processing is complete for each edge, proceed to step S307. If there are any unprocessed edges, return to step S303 and loop.

[0033] Step S307 The score determination unit 33 transmits the determination result and boundary surrounding information to the evaluation result presentation unit 40 and terminates the process. In the process shown in Figure 6, the continuity threshold may be a constant or a ratio to the side length. The boundary surrounding information includes information on the set parameters and coordinate information of the portion exceeding the continuity threshold. This information is necessary for the evaluation result presentation unit 40.

[0034] Figure 7 is a flowchart showing the procedure by which the score determination unit determines the ratio of low-rated portions on each side. The score determination unit 33 sequentially executes the following steps S401 to S407. Step S401: The score determination unit 33 sets the change threshold and the continuity threshold. Then, the process proceeds to step S402. Step S402: The score determination unit 33 obtains a list of sums for each side from the calculation unit 32. Then, the process proceeds to step S403.

[0035] Steps S403 to S406 are loop operations for each edge. In step S403, the score determination unit 33 determines whether the sum value exceeds the change threshold, and if it exceeds it continuously, it calculates the length and coordinate information (boundary information). Then, the process proceeds to step S404. Step S404 The score determination unit 33 determines whether the length of time that continuously exceeds the change threshold exceeds the continuity threshold. If it does not, proceed to step S405. If it does, proceed to step S406. Step S405: The score determination unit 33 sets the percentage (score) of the low-rated portion of the edge in question to "0". If processing is complete for each edge, proceed to step S407. If there are any unprocessed edges, return to step S403 and loop. In step S406, the score determination unit 33 sums the lengths of the portions that exceed the continuity threshold and calculates the ratio (score) to the length of each edge. If processing is complete for each edge, the process proceeds to step S407. If there are any unprocessed edges, the process returns to step S403 and loops.

[0036] Step S407 The score determination unit 33 transmits the determination result and boundary surrounding information to the evaluation result presentation unit 40 and terminates the process. In the process shown in Figure 7, the continuity threshold may be a constant or a ratio to the side length. The boundary surrounding information includes information on the set parameters and coordinate information of the portion exceeding the continuity threshold. This information is necessary for the evaluation result presentation unit 40.

[0037] Figure 8 is a flowchart showing the processing procedure of the parameter search unit. The parameter search unit 50 sequentially executes the following steps S501 to S508. In step S501, the search area setting unit 51 takes evaluated image data (a group of good image data labeled "good" and a group of unacceptable image data labeled "bad") and boundary information as input, and sets the area (parameter range) of the parameters to be searched. After that, the process proceeds to step S502.

[0038] Steps S502 to S507 are loop processes that are executed for each set of parameters. In step S502, the search unit 52 transmits the parameters to the image evaluation unit 30 for setting. Then, the process proceeds to step S503. Step S503: The image evaluation unit 30 performs image evaluation using the set parameters. Then, the process proceeds to step S504. Step S504: The search unit 52 receives the evaluation results (boundary information and score). Then, it proceeds to step S505. In step S505, the search unit 52 changes the decision criteria, draws an ROC (Receiver Operating Characteristic) curve, and calculates AUC (Area Under the Curve). After that, the process proceeds to step S506. Step S506: The search unit 52 determines whether to also search for the judgment criteria. If the judgment criteria are to be searched, the process proceeds to step S507. If the judgment criteria are not to be searched, and processing has been completed for all parameter sets, the process proceeds to step S508. If there are any unprocessed parameter sets, the process returns to step S502 and loops. Step S507: The search unit 52 selects the optimal point from the actual points on the ROC curve. If processing for all parameter sets has been completed, the process proceeds to step S508. If there are any unprocessed parameter sets, the process returns to step S502 and loops. Step S508 The search unit 52 sends the result to the search result presentation unit 53 and terminates the process.

[0039] Figure 9 shows a concrete example of a list display of evaluation results based on the evaluation of each side. Figure 9 shows a list of evaluation results for multiple images, categorized by the number of sides that received a low rating. Specifically, for the image named "2_5_b_0.png" and "4_3_c_0.png", all four sides received a low rating (bad). Also, for the image named "1_0_b_0.png", three sides other than the top side received a low rating (bad). For the image named "1_0_c_0.png", three sides other than the left side received a low rating (bad). Selecting any image name from the list displayed in Figure 9 will switch to a detailed view of the image based on the evaluation of each side.

[0040] Figure 10 shows a specific example of detailed image display based on the evaluation of each side. In Figure 10, the target image is displayed along with the image name "1_0_b_0.png" and the score "3 sides are 'bad'". The target image has its boundaries superimposed as, for example, thin white rectangles, and the low-rated areas are further superimposed as thick white lines. This display allows the user to see which parts the image evaluation system 10 has given a low rating to. For example, if the outline of an object in the target image happens to coincide with the boundary of the generated area, the image evaluation system 10 may judge that area as low-rated. By seeing the image with the boundary and low-rated areas superimposed on the target image, the user can judge whether the judgment of the image evaluation system 10 is appropriate or not.

[0041] Figure 11 shows a specific example of a list display of evaluation results based on the percentage of low-rated portions. Figure 11 associates scores with image names. The scores include the percentage of low-rated areas on the left side, the percentage of low-rated areas on the right side, the percentage of low-rated areas on the top side, the percentage of low-rated areas on the bottom side, and the percentage of low-rated areas in the whole (sum of all four sides). Images can be sorted by specifying any of these criteria. For example, they can be sorted in ascending order of the overall percentage, descending order of the overall percentage, ascending order of the left-side percentage, descending order of the left-side percentage, or any other arbitrary criterion. Selecting any image name from the list in Figure 11 will switch to a detailed view of the image based on the percentage of low-rated portions.

[0042] Figure 12 shows a specific example of detailed image display based on the proportion of low-rated areas. In Figure 12, the target image is displayed along with the image name "1_0_b_0.png" and a score. The score shows the ranking of this image among the image group shown in Figure 11, for each item shown in Figure 11. The target image has its boundaries superimposed as, for example, thin white rectangles, and the low-rated areas are further superimposed as thick white lines. This display allows the user to see which parts the image evaluation system 10 has given a low rating to. For example, if the outline of an object in the target image happens to coincide with the boundary of the generated area, the image evaluation system 10 may judge that part as low-rated. By seeing the image with the boundary and low-rated areas superimposed on the target image, the user can judge whether the judgment of the image evaluation system 10 is appropriate or not.

[0043] Figure 13 shows a specific example of the input reception screen provided by the search area setting unit 51. The search area setting unit 51 displays the input reception screen shown in Figure 13 on the user terminal 20, for example, and accepts input. The input reception screen can accept input such as specifying the storage location for invalid image data, specifying the storage location for good image data, specifying the search range and step size for the parameters to be searched, and specifying the values ​​of parameters not to be searched.

[0044] Figure 14 shows a specific example of a list display of results by the search result presentation unit. In Figure 14, the distance between two points, color space, change threshold, continuity threshold, AUC, and a set of recommended criteria are displayed in a list. The display order can be, for example, in descending order of AUC. The display order may also be reorderable based on any item. Selecting any set from the list view in Figure 14 will take you to a detailed view of the parameter set.

[0045] Figure 15 shows a specific example of a detailed display of the parameter set. Figure 15 displays the parameter set shown in Figure 14 and the judgment results of the image group based on those parameters. The judgment results of the image group are shown in a graph with the vertical axis representing "the percentage of images judged as 'bad' by the image evaluation system 10 among the unacceptable image data labeled 'bad'" and the horizontal axis representing "the percentage of images judged as 'bad' by the image evaluation system 10 among the good image data labeled 'good'". Furthermore, from this graph, the edge threshold, which is the judgment criterion, is set to "3". In other words, it is proposed that if the number of "bad" edges is 3 or more, the image itself should be judged as "bad".

[0046] As described above, the image evaluation system 10 disclosed in the embodiment is characterized by comprising: an information acquisition unit 31 that acquires a target image obtained by applying an image generation process to a partial region of an existing image, boundary information indicating the boundary between the generated region which is the partial region subject to the image generation process and the existing image region which is not subject to the image generation process; a calculation unit 32 that calculates changes in features at the boundary; and a score determination unit 33 which is a determination unit that determines the state of the generated region from the changes in features at the boundary. This configuration and operation allows for evaluation of images that have been partially automatically generated.

[0047] Furthermore, the calculation unit 32 calculates the magnitude of the color change at the boundary, and the determination unit detects the portion where the color change is continuous along the boundary. This configuration and operation allows for the detection of areas where color changes are significant and continuous along the boundary between the generated area and the existing image, in images generated by image generation that create objects in a portion of an existing image. Therefore, it can be used to automatically remove images with unnatural appearances at the boundary between the generated area and the existing image, without user confirmation.

[0048] Furthermore, the calculation unit 32 calculates the amount of change in the color vector inside and outside the boundary. Therefore, it is possible to separate real images or generated images that blend naturally from images with unnatural-looking boundaries.

[0049] Furthermore, the determination unit uses a first threshold for the amount of change in the color vector and a second threshold for the continuity along the boundary while the amount of change in the color vector exceeds the first threshold to determine a score indicating the state of the generation region. Therefore, by comparing the area inside and outside the boundary, it is possible to determine a state where the color change is significant and that the change is continuous along the boundary.

[0050] Furthermore, the configuration may also include an evaluation result presentation unit 40 that presents the evaluation results of the generated region to the user using the score. This configuration allows users to review the entire evaluation result.

[0051] Furthermore, the determination unit designates the portion of the boundary where the change in the color vector exceeds the first threshold and continues for a period of time longer than the second threshold as a low-evaluation portion, and the evaluation result presentation unit 40 superimposes the boundary and the low-evaluation portion onto the target image for visualization and presents it to the user. This configuration and operation allows users to verify individual evaluation results.

[0052] Furthermore, the system is characterized by further comprising a search unit 52 that numerically searches for the optimal values ​​of parameters set by at least one of the information acquisition unit 31, the calculation unit 32, and the determination unit. For example, the search unit 52 repeatedly performs image evaluation using a specified parameter range and evaluated image data to search for the optimal set of parameters. This configuration and operation allows for the automatic search for optimal values ​​for image evaluation parameters.

[0053] Furthermore, the system may also include a search result presentation unit 53 that presents the search results from the search unit to the user. With this configuration and operation, the user can review the parameter search results and determine their appropriateness.

[0054] It should be noted that the present invention is not limited to the embodiments described above, and various modifications are included. For example, the embodiments described above are explained in detail to make the present invention easier to understand, and are not necessarily limited to those having all the configurations described. Furthermore, it is possible to replace or add configurations, not just delete them. For example, in the above embodiment, a point located a predetermined distance inside the boundary was compared with a point located just outside the boundary, but a point located a predetermined distance outside the boundary may also be used for comparison. Furthermore, although the above embodiment focused on the generation region of a rectangle, it can be applied to the generation region of any shape. [Explanation of Symbols]

[0055] 10: Image evaluation system, 11: CPU, 12: Memory, 13: Auxiliary storage device, 14: Network interface, 20: User terminal, 30: Image evaluation unit, 31: Information acquisition unit, 32: Calculation unit, 33: Score determination unit, 40: Evaluation result presentation unit, 50: Parameter search unit, 51: Search area setting unit, 52: Search unit, 53: Search result presentation unit

Claims

1. An information acquisition unit acquires a target image obtained by applying an image generation process to a portion of an existing image, and boundary information indicating the boundary between the generated region, which is the portion of the image generation process that was applied, and the existing image region that was not applied to the image generation process. A calculation unit that calculates the change in characteristics at the boundary, A determination unit that determines the state of the generation region from the change in characteristics at the boundary. An image evaluation system characterized by having the following features.

2. The image evaluation system according to claim 1, The calculation unit calculates the magnitude of the color change at the boundary, The determination unit detects the portion where the color change is continuous along the boundary. An image evaluation system characterized by the following features.

3. The image evaluation system according to claim 2, The image evaluation system is characterized in that the calculation unit calculates the amount of change in the color vector inside and outside the boundary.

4. The image evaluation system according to claim 3, The image evaluation system is characterized in that the determination unit uses a first threshold for the amount of change in the color vector and a second threshold for the continuity along the boundary when the amount of change in the color vector exceeds the first threshold to determine a score indicating the state of the generated region.

5. The image evaluation system according to claim 4, An image evaluation system further comprising an evaluation result presentation unit that presents the evaluation result of the generated region to the user using the score.

6. The image evaluation system according to claim 5, The determination unit determines that the portion of the boundary where the amount of change in the color vector exceeds the first threshold and continues for a period of two thresholds or more is a low-evaluation portion. The evaluation result presentation unit superimposes the boundary and the low-rated portion onto the target image to visualize it and presents it to the user. An image evaluation system characterized by the following features.

7. The image evaluation system according to claim 1, An image evaluation system further comprising a search unit that numerically searches for the optimal value of a parameter set by at least one of the information acquisition unit, the calculation unit, and the determination unit.

8. The image evaluation system according to claim 7, The aforementioned search unit is characterized by repeatedly performing image evaluation using a specified parameter range and evaluated image data to search for the optimal set of parameters.

9. The image evaluation system according to claim 7, An image evaluation system further comprising a search result presentation unit that presents the search results from the search unit to the user.

10. Computers An information acquisition step that acquires a target image obtained by applying an image generation process to a portion of an existing image, and boundary information indicating the boundary between the generated region, which is the portion of the image generation process that was applied, and the existing image region that was not applied to the image generation process. A calculation step for calculating the change in characteristics at the boundary, A determination step of determining the state of the generation region from the change in characteristics at the boundary. An image evaluation method characterized by including the following.

Images