Image discrimination device, image discrimination method, program for image discrimination device, and recording medium storing the program

The image discrimination device automates the classification of 'person' and 'non-person' images using adjustable reference values, enhancing the selection and arrangement process in photo albums.

JP7588257B2Active Publication Date: 2024-11-21FUJIFILM CORP
View PDF 16 Cites 0 Cited by

Patent Information

Application Number
JP2024004361
Authority / Receiving Office
JP · JP
Patent Type
Patents
Current Assignee / Owner
Filing Date
2024-01-16
Publication Date
2024-11-21
Estimated Expiration
2038-03-15

AI Technical Summary

Technical Problem

Selecting and arranging images from a large number of photos taken at events like weddings to effectively represent the atmosphere and chronology is challenging, especially distinguishing between 'person' and 'non-person' images, which often requires manual checking and is not easily automated.

Method used

An image discrimination device and method that sets reference values based on criteria such as human and non-human image ratios, face orientation, and object presence to automatically classify images, allowing users to adjust settings to reflect their intentions.

Benefits of technology

Enables easy and intuitive classification of images as human or non-human, aligning with user intent, and facilitates automated arrangement in albums, reducing manual effort and improving selection accuracy.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 0007588257000001
    Figure 0007588257000001
  • Figure 0007588257000002
    Figure 0007588257000002
  • Figure 0007588257000003
    Figure 0007588257000003
Patent Text Reader

Abstract

To provide an image discrimination device, an image discrimination method, and an image discrimination device program, and a recording medium storing the program capable of relatively easily discriminating between a person image and a non-person image, reflecting the intentions of an album creator, or the like.SOLUTION: A method includes adjusting a position of a slide on a slide bar 52 and setting a reference value for a reference item by a user. When the reference value is set at a position P11, a discrimination object image I31 is determined to be a person image, but discrimination object images I32 and I33 are determined to be non-person images. When the reference value is set at a position P12, the discrimination object images I31 and I32 are determined to be person images, and the discrimination object image I33 is determined to be a non-person image. By changing the reference value setting, it is possible to distinguish between person images and non-person images, reflecting an intention of a user, such as an album creator.SELECTED DRAWING: Figure 7
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to an image discrimination device, an image discrimination method, a program for the image discrimination device, and a recording medium storing the program. [Background technology]

[0002] At events such as weddings, professional photographers may take photos and create albums using the images they have taken. When professional photographers create albums, they select images to be used in the album while considering the intention of the scene and the album story, and judging which images are most appropriate. However, when a large number of images, such as more than 1,000, are taken at one event such as a wedding, it is a heavy burden for the album creator to find the optimal images from among such a large number of images and lay out the album. For this purpose, a method of automatically selecting and editing good photos to be used in an album (Patent Document 1), a method of specifying the state of an image to be inserted into an album template and inserting the specified image (Patent Document 2), a method of detecting whether an image is a failure or not and displaying the detection result (Patent Document 3), a method of extracting a specific area, enlarging it, and determining whether it is in focus (Patent Document 4), etc. have been considered.

[0003] There are also systems that add "people" scenes, "beach" scenes, etc. to images (Patent Document 5), systems that detect people's faces (Patent Document 6), systems that identify areas that are candidates for people regions (Patent Document 7), systems that detect people with high accuracy (Patent Document 8), systems that judge an image to be a landscape photo when there are no people, and judge an image to be a people / commemorative photo or portrait image when people are present based on the number of people, area, etc. (Patent Document 8). [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Patent No. 5655112 [Patent Document 2] Patent No. 4455302 [Patent Document 3] JP 2002-187329 A [Patent Document 4] Patent No. 4346859 [Patent Document 5] JP 2007-295212 A [Patent Document 6] JP 2006-318034 A [Patent Document 7] JP 2007-264887 A [Patent Document 8] JP 2008-288684 A Summary of the Invention [Problem to be solved by the invention]

[0005] When creating an album, not only "portrait images" but also "non-portrait images" such as food, buildings, and landscapes that enhance the memories are often selected and laid out in the album. In particular, when a professional photographer creates an album using images taken at an event such as a wedding, it is important to select "non-portrait images" that effectively express the atmosphere of the wedding in addition to "portrait images" such as the bride and groom. These "non-portrait images" are not necessarily arranged in chronological order according to the flow of the event such as a wedding, so it can be difficult to select from the many images taken. In this case, it is possible to distinguish between "portrait images" and "non-portrait images" from among the many images and find the "non-portrait images" to be used in the album. However, what kind of images are "portrait images" and what kind of images are "non-portrait images" will vary depending on the intention of the album creator and the nature of the captured image (such as the angle of view intended by the photographer). For this reason, album creators often have to check a large number of images one by one to determine whether they are "images of people" or "images of non-people," and it is often not possible to easily distinguish between "images of people" and "images of non-people" in a way that reflects the album creator's intentions.

[0006] The object of the present invention is to relatively easily distinguish between "portrait images" and "non-portrait images" while reflecting the intention of the album creator, etc. [Means for solving the problem]

[0007] The image discrimination device of the present invention is characterized by comprising a reference value setting means for setting a reference value to be used in the process of discriminating at least one of a human image and a non-human image for at least one criterion item, and a discrimination means for discriminating whether an image to be discriminated is a human image or a non-human image (a non-human image is an image that is not a human image) based on the reference value set by the reference value setting means.

[0008] The present invention also provides an image discrimination method suitable for an image discrimination device, in which a reference value setting means sets a reference value used in a process of discriminating at least one of a human image and a non-human image for at least one criterion item, and a discrimination means discriminates whether an image to be discriminated is a human image or a non-human image based on the set reference value.

[0009] The present invention also provides a computer-readable program for controlling the computer of the image discrimination device, and a recording medium storing the program.

[0010] In addition, an image discrimination device may be provided that sets a reference value for use in the process of discriminating between at least one of a human image and a non-human image for at least one criterion item, and has a processor that discriminates whether the image to be discriminated is a human image or a non-human image based on the set reference value.

[0011] The discrimination means discriminates whether the image to be discriminated is a human image or a non-human image based on an evaluation value indicating the likeliness of the image to be discriminated as a human image or a non-human image, which is obtained, for example, based on a plurality of sample images, and a reference value set by a reference value setting means.

[0012] The criterion item may be, for example, an area ratio of a human part or a non-human part in the discrimination target image, and the discrimination device may further include a detection means for detecting the human part and the non-human part in the discrimination target image. In this case, the discrimination means determines whether the discrimination target image is a human image or a non-human image based on, for example, the area ratio of the human part or the non-human part in the discrimination target image detected by the detection means and the reference value set by the reference value setting means.

[0013] The image processing device may further comprise a display control means for displaying the discrimination target image by changing the display manner of the human portion and the non-human portion in the discrimination target image detected by the detection means.

[0014] The discrimination means may, for example, determine whether the image to be discriminated is an image of a person or an image of a non-person based on the direction of the person's face in the image to be discriminated and a reference value set by the reference value setting means.

[0015] The criteria may be, for example, the ratio of people to non-people objects contained in the image to be distinguished, the ratio of people to landscapes contained in the image to be distinguished, or the degree to which the faces of people contained in the image to be distinguished are facing forward.

[0016] When there are multiple criterion items, the discrimination means may, for example, discriminate the image to be discriminated for each reference value for each criterion item set by multiple reference value setting means, and output a discrimination result for each reference value.

[0017] Based on the discrimination results for a plurality of criteria, the image to be discriminated may be discriminated as being either a human image or a non-human image.

[0018] The device may further include an alarm means for alarming the discrimination result of the discrimination means, and an alarm control means for changing the alarm mode of the alarm means in response to a change in the discrimination result of the discrimination means due to a change in the setting of the reference value in the reference value setting means.

[0019] The image processing device may further comprise an adoption decision means for deciding whether or not to adopt the image to be discriminated based on the discrimination result by the discrimination means.

[0020] The image processing device may further comprise a position determining means for determining a position in the album where the discrimination target image determined to be adopted by the adoption determining means is to be pasted based on the discrimination result of the discrimination means. Effect of the Invention

[0021] According to this invention, by changing the reference value in accordance with the user's intention, the same image to be discriminated can be discriminated as a human image or a non-human image based on the user's intention, so that it is possible to determine whether the image to be discriminated is a human image or a non-human image, reflecting the intention of the user, such as the album creator. [Brief description of the drawings]

[0022] [Figure 1] FIG. 2 is a block diagram showing an electrical configuration of the image discrimination device. [Diagram 2] FIG. 2 is an example of a block diagram showing the electrical configuration of an artificial intelligence device. [Diagram 3] 1 is an example of a sample image stored in a database. [Figure 4] 4 is a flowchart showing a processing procedure of the image discrimination device. [Diagram 5] 2 is an example of an image display window. [Figure 6] 13 is an example of a reference value setting window. [Figure 7] The relationship between the image to be discriminated and the slide bar is shown. [Figure 8] The relationship between the image to be discriminated and the slide bar is shown. [Figure 9] The relationship between the image to be discriminated and the slide bar is shown. [Figure 10] 1 is an example of an image classification window. [Figure 11] This is an example of a double-page spread of an album. [Figure 12] 13 is a flowchart showing a procedure of a determination process. [Figure 13] 13 is a flowchart showing a procedure of a determination process. [Figure 14] Non-human parts and human parts are detected. [Figure 15] 1 is an example of an image. [Figure 16] 1 is an example of an image. [Figure 17] 13 is a flowchart showing a procedure of a determination process. [Figure 18] 13 is a flowchart showing a procedure of a determination process. [Figure 19] 13 is an example of a reference value setting window. [Figure 20] 13 is an example of a determination result display window. [Figure 21] 4 is a flowchart showing a processing procedure of the image discrimination device. [Figure 22] 4 is a flowchart showing a processing procedure of the image discrimination device. [Diagram 23] 13 is an example of a determination result display window. [Figure 24] 1 is a part of a flowchart showing a processing procedure of the image discrimination device. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0023] [First Example] FIG. 1 is a block diagram showing the electrical configuration of an image discrimination device 1 according to an embodiment of the present invention.

[0024] The overall operation of the image discrimination device 1 is controlled by a CPU (Central Processing Unit) 2 .

[0025] Image discrimination device 1 includes a display device 3 that displays images and other information on a display screen, and a communication device 4 that connects to the Internet or other networks and communicates with devices other than image discrimination device 1. Image discrimination device 1 also includes a hard disk 5, a hard disk drive 6 that accesses hard disk 5, a memory 7 that stores data, etc., and a keyboard 8 and mouse 9 that input commands, etc. Image discrimination device 1 also includes a compact disk drive 10 that accesses compact disk 11, and a memory card reader / writer 12 that writes data to memory card 13 and reads data recorded on memory card 13.

[0026] An operation program for image discrimination device 1, which will be described later, is received by communication device 4 via the Internet. The received operation program is installed in image discrimination device 1. The operation program may be received by image discrimination device 1 via a network such as the Internet and not installed in image discrimination device 1, but recorded on a portable recording medium such as compact disc 11 and read from the portable recording medium. In that case, the operation program read from the portable recording medium is installed in image discrimination device 1. It goes without saying that the operation program is readable by CPU 2 (computer) of image discrimination device 1.

[0027] Fig. 2 is a block diagram showing the electrical configuration of an artificial intelligence device 20 capable of communicating with the image discrimination device 1. In Fig. 2, it is described as being configured by hardware, but it goes without saying that a part of it may be configured by software.

[0028] The artificial intelligence device 20 includes a database 21 .

[0029] FIG. 3 shows a sample image I represented by a number of image files stored in a database 21 .

[0030] A score of the degree of personhood is assigned to each sample image I for each criterion item in advance. In this embodiment, a score of the degree of personhood based on the ratio of people to "objects / items" (one example of a non-human part) in the sample image I (for example, the higher the ratio of people, the higher the score), a score of the degree of personhood based on the ratio of people (human parts) to "landscape" (one example of a non-human part) in the sample image I (for example, the higher the ratio of people, the higher the score), and a score of the degree of personhood based on the direction of the person's face in the sample image I (for example, the more the person's face faces forward, the higher the score) are assigned to each sample image I. Data representing the scores is stored, for example, in the header of an image file representing the sample image I. The degree of personhood represents an index, degree, and evaluation value of the likeness of the sample image I as a human image, and the higher the score (for example, a maximum of 3 points, but other scores are also acceptable), the more likely it is to be a human image. If the score of the degree of personhood is low, the image in the sample image I is more likely to be a non-human image. Data representing a score of non-human degree (representing an index, degree of indicating that sample image I is a non-human image, and an evaluation value of the likeliness of it being a non-human image) instead of a score of human degree may be stored in the header of the image file representing sample image I, or data representing the score of human degree and the score of non-human degree may be stored in the header of the image file representing sample image I. The user will assign a score of human degree or non-human degree while viewing the image.

[0031] 2, a large number of image files stored in a database 21 are read, and data representing image features is extracted in a feature extraction device 22. A learning model 24 is generated according to a learning algorithm 23 using the data representing the extracted features.

[0032] Image data is transmitted from communication device 4 of image discrimination device 1 and input from communication device 25 of artificial intelligence device 20. The input image data is used in learning model 24 to calculate a score of the person degree (one example of an evaluation value representing the likeliness of an image being a person image) (a score of the non-person degree may also be used. The non-person degree score is one example of an evaluation value representing the likeliness of an image being a non-person image) of the image represented by the image data. Data representing the calculated score is transmitted from communication device 25 to image discrimination device 1.

[0033] When the data representing the score transmitted from the artificial intelligence device 20 is received by the image discrimination device 1, the CPU 2 determines, based on the score, whether the image represented by the image data transmitted from the image discrimination device 1 is an image of a person or a non-person image.

[0034] The artificial intelligence device 20 shown in FIG. 2 is a device separate from the image discrimination device 1 shown in FIG. 1, but may be connected to the image discrimination device 1 shown in FIG.

[0035] FIG. 4 is an example of a flowchart showing a processing procedure of the image discrimination device 1.

[0036] When a command such as an image discrimination start command is given to image discrimination device 1 from keyboard 8 or the like, the process shown in FIG. 4 starts.

[0037] It is assumed that image files representing a large number of images to be determined as human or non-human images are stored in memory card 13. Of course, image files representing a large number of images to be determined as human or non-human images may be stored in a recording medium other than memory card 13, such as compact disk 11 or hard disk 5.

[0038] A large number of image files stored in the memory card 13 are read out by the memory card reader / writer 12, and a large number of images are displayed in a list on the display screen of the display device 3 (step 31).

[0039] FIG. 5 is an example of an image display window 40 .

[0040] The image display window 40 is displayed on the display screen of the display device 3. An image display area 41 is formed in the display window 40. A large number of discrimination target images I1 to I28 represented by a large number of image files stored in the memory card 13 are displayed in this image display area 41. A scroll bar 42 is formed on the right side of the image display area 41. A knob 43 is formed on the scroll bar 42. The knob 43 can be freely moved up and down, and when the user moves the knob 43 up and down, an image that is not displayed in the image display area 41 is displayed in the image display area 41.

[0041] Next, the user sets a reference value (step 32 in FIG. 4).

[0042] FIG. 6 is an example of the reference value setting window 50.

[0043] A reference value setting window 50 shown in FIG.

[0044] In this embodiment, the user can set a reference value (threshold value) used in the process of determining whether an image is a human image or a non-human image for any one of multiple reference items. For example, as described with reference to Fig. 2, if the score calculated in the artificial intelligence device 20 and transmitted from the artificial intelligence device 20 to the image discrimination device 1 is equal to or greater than the reference value, the image discrimination device 1 determines that the image with that score is a non-human image (or a human image). Since the user can set the reference value, it is possible to determine whether an image is a non-human image or a human image according to the reference value set by the user. It becomes possible to determine whether an image is a non-human image or a human image by reflecting the user's intention, etc.

[0045] In the reference value setting window 50, a first check box 51, a second check box 61, and a third check box 71 are formed.

[0046] To the right of the first check box 51, a character string "Ratio of people to 'things and items'" and a first horizontal slide bar 52 (an example of a reference value setting means) are displayed. A first slider 53 that can be moved horizontally is formed on the first slide bar 52. The first reference item is the "Ratio of people to 'things and items'". The reference value of the first reference item is set according to the position of the first slider 53 on the first slide bar 52. The more the position of the first slider 53 is positioned to the left, the higher the score of the human-likeness of the image to be discriminated (when the proportion of people is higher than the proportion of things and items contained in the image to be discriminated) is determined to be a human image, and the more the position of the first slider 53 is positioned to the right, the lower the proportion of people contained in the image to be discriminated is determined to be a human image. The ratio of people to "things and items" refers to the ratio of the parts of people contained in the image to be discriminated to the things and items (goods) held by the person, excluding the background parts of the person, such as scenery. If an object or the like is included in the background of a person, it is ignored, but the object included in the image to be discriminated may be detected and the ratio of the person to the object may be calculated.

[0047] Figure 7 shows how the image being classified is judged to be a human or non-human image based on the ratio of people to "objects / items."

[0048] In discrimination target image I31, the proportion of the object (bouquet) is small compared to the size of the person, and the person-degree score of discrimination target image I31 is 3 points. In addition, in discrimination target image I32, the proportion of the object (bouquet) is relatively large compared to the size of the person, and the person-degree score of discrimination target image I32 is 2 points. In addition, in discrimination target image I33, the proportion of the object (bouquet) is large compared to the size of the person, and the score of discrimination target image I33 is 1 point.

[0049] In such a case, for example, when the first slider 53 is positioned at position P11 of the first slide bar 52 and set to discriminate images with a score of 2.5 or more as human images and images with a score of less than 2.5 as non-human images, the discrimination target image I31 is discriminated as a human image, but the discrimination target images I32 and I33 are discriminated as non-human images. However, when the first slider 53 is positioned at position P12 of the first slide bar 52 and set to discriminate images with a score of 1.5 or more as human images and images with a score of less than 1.5 as non-human images, the discrimination target images I31 and I32 are discriminated as human images, and the discrimination target image I33 is discriminated as a non-human image. In this way, the discrimination target image I32 can be discriminated as a human image or a non-human image according to the user's intention (according to the position of the first slider 53). 7 are displayed in correspondence with the first slide bar 52, so that it is possible to know what kind of images are judged to be human images or non-human images according to the position of the first slider 53. Also, a plurality of sample images may be displayed, and the user may specify the sample images as human images or non-human images, and the reference value may be determined based on the human-image-likeness score or non-human image-likeness score of the specified human image or non-human image.

[0050] Returning to FIG. 6, the character string "Ratio of People to Landscape" and a second horizontal slide bar 62 (an example of a reference value setting means) are displayed on the right side of the second check box 61. A second slider 63 that can be moved horizontally is also formed on the second slide bar 62. The second reference item is the "Ratio of People to Landscape". The reference value of the second reference item is set according to the position of the second slider 63 on the second slide bar 62. The more the position of the second slider 63 is positioned to the left, the higher the score of the person-likeness of the image to be discriminated (the more people there are compared to the landscape contained in the image to be discriminated), and the more the position of the second slider 63 is positioned to the right, the lower the ratio of people contained in the image to be discriminated. The ratio of people to landscape is the ratio of the part of the person contained in the image to be discriminated to the landscape (background). The landscape (background) may be regarded as the part other than the area surrounded by a person. Even if a person in the target image is holding an object such as a bouquet, the object is ignored. Objects that exist in the landscape (background) are considered to be part of the landscape, but objects may be detected from the target image and the ratio may be calculated using the landscape (background) excluding the detected objects.

[0051] In the above-described embodiment, a distinction is made between scenery (background) and objects, but it is also possible to detect the human parts and non-human parts (parts other than people) contained in the target image without distinguishing between scenery (background) and objects, and set a reference value using the ratio of the human parts to the non-human parts as a criterion item.

[0052] FIG. 8 shows how the image being classified is judged to be a human image or a non-human image based on the ratio of people to "landscape."

[0053] Assume that the proportion of scenery (background) in discrimination target image I41 is small compared to the size of the person, and the human degree score for discrimination target image I41 is 3 points. Assume also that the proportion of scenery in discrimination target image I42 is relatively large compared to the size of the person, and the human degree score for discrimination target image I42 is 2 points. Assume also that the proportion of scenery in discrimination target image I43 is large compared to the size of the person, and the score for discrimination target image I43 is 1 point.

[0054] In such a case, for example, when the second slider 63 is positioned at position P21 of the second slide bar 62 and set to discriminate images with a score of 2.5 or more as human images and images with a score of less than 2.5 as non-human images, the discrimination target image I41 is discriminated as a human image, but the discrimination target images I42 and I43 are discriminated as non-human images. However, when the second slider 63 is positioned at position P21 of the second slide bar 62 and set to discriminate images with a score of 1.5 or more as human images and images with a score of less than 1.5 as non-human images, the discrimination target images I41 and I42 are discriminated as human images, and the discrimination target image I43 is discriminated as a non-human image. In this way, the discrimination target image I42 can be discriminated as a human image or a non-human image according to the user's intention (according to the position of the second slider 63). 8 are displayed in correspondence with the second slide bar 62, so that it is possible to know what kind of images are judged to be human images or non-human images according to the position of the second slider 63. Also, a plurality of sample images may be displayed, and the user may specify the sample images as human images or non-human images, and the reference value may be determined based on the human-image-likeness score or non-human image-likeness score of the specified human image or non-human image.

[0055] Returning to FIG. 6, to the right of the third check box 71, the character string "human face direction degree" and a third horizontal slide bar 72 (one example of a reference value setting means) are displayed. A third slider 73 that is freely movable in the horizontal direction is also formed on the third slide bar 72. The third reference item is "human face direction degree". The reference value of the third reference item is set according to the position of the third slider 73 on the third slide bar 72. The further to the left the position of the third slider 73 is positioned, the higher the human-likeness score of the image to be discriminated is (the image to be discriminated contains a higher proportion of people compared to the scenery), and the further to the right the position of the third slider 73 is positioned, the lower the proportion of people in the image to be discriminated is determined to be a human image.

[0056] FIG. 9 shows how to determine whether an image to be discriminated is a human image or a non-human image based on the degree of face direction of a person.

[0057] In the discrimination target image I51, a large proportion of the people's faces are facing forward, and the person degree score for discrimination target image I51 is 3 points. In addition, in the discrimination target image I52, a relatively small proportion of the people's faces are facing forward, and the person degree score for discrimination target image I52 is 2 points. Furthermore, in the discrimination target image I53, the people's faces are not facing forward, and the score for discrimination target image I53 is 1 point.

[0058] In such a case, for example, when the third slider 73 is positioned at position P21 of the third slide bar 72 and set to discriminate images with a score of 2.5 or more as human images and images with a score of less than 2.5 as non-human images, the discrimination target image I51 is discriminated as a human image, but the discrimination target images I52 and I53 are discriminated as non-human images. However, when the third slider 73 is positioned at position P21 of the third slide bar 72 and set to discriminate images with a score of 1.5 or more as human images and images with a score of less than 1.5 as non-human images, the discrimination target images I51 and I52 are discriminated as human images, and the discrimination target image I53 is discriminated as a non-human image. In this way, the discrimination target image I52 can be discriminated as a human image or a non-human image according to the user's intention (according to the position of the third slider 73). 8 are displayed in correspondence with the third slide bar 72, so that it is possible to know what kind of images are judged to be human images or non-human images according to the position of the third slider 73. Also, a plurality of sample images may be displayed, and the user may specify the sample images as human images or non-human images, and the reference value may be determined based on the scores of the specified human images or non-human images as to their likeness as human images or non-human images.

[0059] Referring to FIG. 6, when any of the first check box 51, the second check box 61 and the third check box 71 is checked (it is also possible to check multiple check boxes as described later), the position of the first slider 53, the second slider 63 or the third slider 73 of the first slide bar 52, the second slide bar 62 or the third slide bar 72 formed corresponding to the checked check box is set as a reference value.

[0060] After the reference value is set, when a command to start discrimination is input to the image discrimination device 1 by the keyboard 8 or the like (YES in step 33 in FIG. 4), image data representing the first discrimination target image is read from the memory card 13 (step 34 in FIG. 4). A process of discriminating whether the image represented by the read image data is a human image or a non-human image is performed by the CPU 2 (an example of a discrimination means) based on the set reference value (step 35 in FIG. 4). This discrimination process will be described later. A process of discriminating whether the discrimination target image is a human image or not may be performed without discriminating whether the discrimination target image is a human image or not, and if it is not discriminated as a human image, it may be discriminated as a non-human image, or a process of discriminating whether the discrimination target image is a non-human image may be performed, and if it is not discriminated as a non-human image, it may be discriminated as a human image. Depending on the discrimination result, the discrimination target image is classified and displayed (step 36 in FIG. 4).

[0061] 10 is an example of an image classification window 80. The image classification window 80 is also displayed on the display screen of the display device 3 of the image discrimination device 1.

[0062] The image classification window 80 includes a person image window 81 and a non-person image window 82. The person image window 81 displays discrimination target images I1, I2, and I4-I6 that have been determined to be person images from among the images to be discriminated. The non-person image window 82 displays discrimination target images I3 and I7 that have been determined to be non-person images from among the images to be discriminated. The user can tell that the image displayed in the person image window 81 is a person image, and that the image displayed in the non-person image window 82 is a non-person image.

[0063] Next, if there is a next image to be discriminated (YES in step 37 in FIG. 4), the processing in steps 34 to 36 is repeated, and image data representing the next image to be discriminated is read from the memory card 13 (step 34 in FIG. 4).

[0064] FIG. 11 is an example of a double-page spread that makes up an album.

[0065] Facing page 90 constituting the album shown in Fig. 11 is also displayed on the display screen of display device 3 of image discrimination device 1, but may also be displayed on the display screen of a display device other than image discrimination device 1. Although the album contains pages other than facing page 90, only facing page 90 is shown in Fig. 11.

[0066] A plurality of image insertion areas 91-96 are formed on a double-page spread 90 of the album. The user places human images and non-human images in these image insertion areas 91-96 while taking into consideration the album story. For example, a discrimination target image I6 of the bride and groom before the cake is cut is inserted in image insertion area 91, and a discrimination target image I7 of the cake is inserted in image insertion area 92. After that, if there is an image of the cake cutting scene, for example, that image is inserted in image insertion area 93. Human images and non-human images may be automatically placed without the user placing images.

[0067] FIG. 12 is a flowchart showing the first determination processing procedure (the processing procedure of step 36 in FIG. 4).

[0068] The first discrimination process procedure uses the artificial intelligence device 20 as described above.

[0069] Image data representing the discrimination target image read from the memory card 13 and data representing the criterion item selected from the first check box 51, the second check box 61 and the third check box 71 are transmitted from the image discrimination device 1 to the artificial intelligence device 20 (step 101). In the artificial intelligence device 20, a score of the likeliness of the discrimination target image being a human image for that criterion item is calculated using the score given for each criterion item to the images (which are examples of sample images) stored in the database 21 according to the selected criterion item. Data representing the calculated score (evaluation value of the likeliness of being a human image) is transmitted from the artificial intelligence device 20 to the image discrimination device 1. The data representing the score transmitted from the artificial intelligence device 20 is received by the communication device 4 of the image discrimination device 1 (step 102).

[0070] Based on the received data representing the score and the reference values ​​of the criteria items set by the user, the image to be discriminated is discriminated as a human image or a non-human image (step 103). For example, if the score represented by the received data is equal to or greater than the set reference value, the image is discriminated as a human image by the CPU 2, and if the score is less than the reference value, the image is discriminated as a non-human image by the CPU 2.

[0071] FIG. 13 is a flowchart showing the second determination process procedure (the process procedure of step 35 in FIG. 4).

[0072] The processing procedure shown in FIG. 13 is performed when the first criterion item or the second criterion item is set.

[0073] The human and non-human parts of the image to be discriminated are detected by the CPU 2 (detection means) of the image discrimination device 1 (step 111). When the first criterion item is set as described above, "objects / items" are detected as non-human parts, and when the second criterion item is set, "landscape" parts (background parts) are detected as non-human parts.

[0074] For example, as shown in FIG. 14, when the image to be discriminated is image I5 and the second criterion item is set, a "landscape" portion (background portion) 121 is detected as a non-human portion, and a human portion 122 is detected.

[0075] Returning to FIG. 13, the area ratio between the non-human portion 121 and the human portion 122 is calculated (step 112). From the calculated area ratio and the set reference value, it is determined whether the image to be discriminated is a human image or a non-human image (step 113). For example, the left end of the second slide bar 62 of the second criterion item corresponds to the large human degree so that the area ratio of the "human portion" to the "non-human portion" is 1:2, and the right end of the second slide bar 62 of the second criterion item corresponds to the large non-human degree so that the area ratio of the "human portion" to the "non-human portion" is 1:10 (or only non-human). If the calculated area ratio corresponds to a position to the left of the ratio of the "human portion" to the "non-human portion" determined by the position of the second slider 63, the image to be discriminated is a human image, and if the calculated area ratio corresponds to a position to the right, the image to be discriminated is a non-human image. For example, if the second slider 63 is positioned at the center of the second slide bar 62, and the center position indicates that the area ratio of "human parts" to "non-human parts" is 1:6, then if the area ratio of the image to be discriminated is such that the human parts are more numerous than 1:6, then the image is determined to be a human image, and if the area ratio of the image to be discriminated is such that the non-human parts are more numerous than 1:6, then the image is determined to be a non-human image. The same applies when the first criterion item is set instead of the second criterion item.

[0076] 15 and 16 show an example in which the display mode of the discrimination target image I5 is changed.

[0077] Fig. 15 is an example of an image 131 in which the landscape (background) portion 121 in the discrimination target image I5 is darkened compared to the person portion 122. The non-person portion 121 detected as described in Fig. 14 is displayed darkened by the CPU 2 (this is an example of changing the display manner of the person portion and the non-person portion). The darkened portion is indicated by hatching.

[0078] Fig. 16 is an example of an image 132 in which the person portion 122 in the discrimination target image I5 is darkened compared to the scenery (background) portion 121. As explained in Fig. 14, the detected person portion 122 is displayed darkened by the CPU 2 (this is an example of changing the display manner of the person portion and the non-person portion). The darkened portion is indicated by hatching.

[0079] In this way, the CPU 2 (display control means) may change the display manner of the human portion 122 and the non-human portion 121 in the discrimination target image and display them on the display screen of the display device 3. The human portion 122 and the non-human portion 121 may be displayed in different colors without changing the brightness.

[0080] FIG. 17 is a flowchart showing the third determination process procedure (the process procedure of step 35 in FIG. 4).

[0081] The process shown in Fig. 17 is performed when the third check box 71 is checked in the reference value setting window 50 shown in Fig. 6 to select the third reference item. First, the CPU 2 detects a face from within the image to be discriminated (step 141). Then, the CPU 2 detects the orientation of the detected face (step 142). The orientation of the face can be detected from the position of the eyes relative to the face, the distance between the eyes, the direction of the nose, the size of the mouth, and the like. Based on the detected orientation of the face and the reference value set by the third slider 73 of the third slide bar 72, it is determined whether the image to be discriminated is a human image or a non-human image (step 143).

[0082] For example, the left end of the third slide bar 72 shown in Fig. 6 indicates a face facing directly forward, and the right end of the third slide bar 72 indicates a face facing backwards. The third slider 73 is positioned in the center, specifying a position where the face is facing sideways. In such a case, if the calculated face direction is facing forward rather than sideways, the image to be discriminated is determined to be a human image, and if the calculated face direction is facing backwards rather than sideways, the image to be discriminated is determined to be a human image.

[0083] As described above, with regard to the direction of the face, data representing the image to be discriminated may be transmitted to the artificial intelligence device 20, and the artificial intelligence device 20 may calculate a score based on the direction of the face contained in the image to be discriminated, and may determine whether the image to be discriminated is a human image or a non-human image based on the calculated score and a reference value determined by the position of the third slider 73.

[0084] According to the first embodiment, it becomes possible to distinguish between human images and non-human images according to the user's intention.

[0085] [Second Example] 18 to 20 show an embodiment in which it is determined whether an image to be discriminated is a human image or a non-human image based on a plurality of criteria items.

[0086] FIG. 18 is a flowchart showing the fourth determination process procedure (the process procedure of step 35 in FIG. 4).

[0087] In this process, a plurality of check boxes can be checked among the first check box 51, the second check box 61, and the third check box 71 included in the reference value setting window 50. In this embodiment, it is assumed that the first check box 51, the second check box 61, and the third check box 71 are all checked, and the reference values ​​of the ratio of a person to "objects / items" is set by the first slider 53, the ratio of a person to "landscape" by the second slider 63, and the degree of a person's face direction by the third slider 73.

[0088] Image data representing the image to be discriminated is read, and the CPU 2 detects the human and non-human parts contained in the image to be discriminated (step 161). Because the first check box 51 is checked, the "objects / items" parts are detected as non-human parts, and because the second check box 61 is also checked, the "landscape" parts are also detected as non-human parts. When the human and non-human parts of the image to be discriminated are detected, the area ratio between the human and non-human parts is calculated by the CPU 2 (step 162). In calculating this area ratio, a first area ratio between the human parts and the "objects / items" parts and a second area ratio between the human parts and the "landscape" parts are respectively obtained.

[0089] Next, a face is detected by the CPU 2 from within the image to be discriminated (step 163).

[0090] Next, the CPU 2 calculates an individual discrimination value for the area ratio from each of the calculated first and second area ratios and a set reference value (step 165). The individual discrimination value calculated from the area ratio and the set reference value indicates the difference between the area ratio and the reference value, and includes a first individual discrimination value calculated from a first area ratio between the person portion and the "object / item" portion and a reference value set by the slider 53, and a second individual discrimination value calculated from a second area ratio between the person portion and the "landscape" portion and a reference value set by the slider 63. Also, the CPU 2 calculates an individual discrimination value for the face direction (hereinafter referred to as a third individual discrimination value) from the face direction and a set reference value (step 166). The third individual discrimination value calculated from the face direction and the reference value indicates the difference between the detected face direction and the reference value. These first to third individual discrimination values ​​are calculated as follows.

[0091] FIG. 19 shows the reference value setting window 50 .

[0092] The value of the first area ratio between the person portion and the "things and items" portion corresponds to the position P51 on the first slide bar 52 (for example, data representing the first area ratio is sent to the artificial intelligence device 20, and the score of the degree of personness at the first area ratio is calculated based on a large number of sample images I stored in the database 21. The position corresponding to the calculated score of the degree of personness is the position P51), and the first slider 53 is set by the user to the position P50. The distance between the positions P50 and P51 is represented by L1. If the position P51 is to the left of the position P50 of the first slider 53, the image to be discriminated is discriminated as a person image in the discrimination based on the first area ratio between the person portion and the "things and items" portion, and if the position P51 is to the right of the position P50 of the first slider 53, the image to be discriminated is discriminated as a non-person image in the discrimination based on the first area ratio between the person portion and the "things and items" portion. In the example shown in Fig. 19, since the position of P51 is to the left of the first slider 53, the discrimination target image is discriminated as a human image in discrimination based on the first area ratio between the person portion and the "object / item" portion (this is an example of a discrimination result for each criterion value in the criterion item), and the distance L1 is a positive distance. The value of this distance + L1 corresponds to the first individual discrimination value.

[0093] Similarly, the value of the second area ratio between the person portion and the "landscape" portion corresponds to the position P61 on the second slide bar 62 (for example, data representing the second area ratio is sent to the artificial intelligence device 20, and the score of the person degree at the second area ratio is calculated based on a large number of sample images I stored in the database 21. The position corresponding to the calculated person degree score is the position P61), and the second slider 63 is set by the user to the position P60. The distance between the positions P60 and P61 is represented by L2. If the position P61 is to the left of the position P60 of the second slider 63, the image to be discriminated is a person image in the discrimination based on the second area ratio between the person portion and the "landscape" portion, and if the position P61 is to the right of the position P60 of the second slider 63, the image to be discriminated is a non-person image in the discrimination based on the second area ratio between the person portion and the "landscape" portion. Since the position of P61 is to the right of the second slider 63, it is determined to be a non-human image in the determination based on the second area ratio between the human portion and the "landscape" portion (this is an example of a determination result for each criterion value in the criterion item), and the distance L2 is set to a negative distance. This value of the distance -L2 corresponds to the second individual determination value.

[0094] Similarly, the face direction value corresponds to the position P71 on the third slide bar 72, and corresponds to the position P70 on the third slider 73, and the third slider 73 is set by the user to the position P70. The distance between the positions P70 and P71 is represented by L3. If the position of P71 is to the left of the position P70 on the third slider 73, the discrimination based on the face direction discriminates the image to be discriminated as a person image, and if the position of P71 is to the right of the position P70 on the third slider 73, the discrimination based on the face direction discriminates the image to be discriminated as a non-person image. Since the position of P71 is to the left of the third slider 73, the discrimination based on the face direction discriminates the image to be discriminated as a person image (this is an example of a discrimination result for each reference value in the criterion item), and the distance L3 is a positive distance. The value of this distance + L3 corresponds to the third individual discrimination value.

[0095] From the first individual discrimination value, the second individual discrimination value, and the third individual discrimination value thus obtained, an overall discrimination value is calculated, and the CPU 2 discriminates whether the discrimination target image is a person image or a non-person image (step 167). For example, the sum of the first individual discrimination value, the second individual discrimination value, and the third individual discrimination value may be used as the overall discrimination value (if the sum is positive, the image is discriminated as a person image, and if the sum is negative, the image is discriminated as a non-person image), or a priority may be set and the first individual discrimination value, the second individual discrimination value, and the third individual discrimination value may be weighted according to the priority, and the weighted sum may be used as the overall discrimination value.

[0096] FIG. 20 shows an example of the discrimination result display window 150. As shown in FIG.

[0097] A discrimination result display window 150 is also displayed on the display screen of the display device 3 of the image discrimination device 1. The discrimination result display window 150 displays the discrimination target image I5. To the right of this discrimination target image I5, text 151 of the discrimination result based on the first individual discrimination value, text 152 of the discrimination result based on the second individual discrimination value, text 153 of the discrimination result based on the third individual discrimination value, and text 154 of the discrimination result based on the overall discrimination value are displayed (an example of output of discrimination results for a plurality of criteria items). The user can not only know whether the discrimination target image I5 has been comprehensively discriminated as a human image or a non-human image, but also know the discrimination results based on each of the first individual discrimination value, the second individual discrimination value, and the third individual discrimination value.

[0098] According to the second embodiment, it is possible to know whether the image to be discriminated is a human image or a non-human image based on a plurality of criterion items. Moreover, the discrimination results based on each of the plurality of criterion items are also known.

[0099] [Third Example] Figures 21 to 23 show another embodiment in which the reference value is changed after a discrimination result indicating whether a plurality of images to be discriminated are human images or non-human images is obtained.

[0100] Figures 21 and 22 are flowcharts showing the processing procedure of the image discrimination device 1. In Figures 21 and 22, the same processes as those shown in Figure 4 are denoted by the same reference numerals.

[0101] A plurality of images to be discriminated are displayed on the display screen of the display device 3 (step 31), and a reference value (which may be one or more) is set by the user using the reference value setting window 50 shown in Fig. 6 (step 32). When a command to start discrimination is input (YES in step 33), the images to be discriminated are read (step 34), and discrimination processing is performed (step 35). As shown in Fig. 10, the images to be discriminated are classified and displayed according to the discrimination results (step 36). If an image to be discriminated exists (YES in step 37), it is confirmed whether the reference value setting has been changed (step 171 in Fig. 22).

[0102] When a checked box among the first check box 51, the second check box 61 and the third check box 71 included in the reference value setting window 50 is changed, or when the position of any of the sliders among the first slider 53, the second slider 63 and the third slider 73 corresponding to the checked box is changed, the CPU 2 determines that the reference value setting has been changed.

[0103] When the reference value setting is changed (YES in step 171), image data representing an already-discriminated discrimination target image is read again (step 172), and discrimination processing is performed again based on the changed reference value (step 173). When the discrimination result changes due to the change in the reference value (YES in step 174), a frame is attached to the discrimination target image whose discrimination result has changed (step 175).

[0104] FIG. 23 is an example of an image classification window 80.

[0105] The image classification window 80 includes a person image window 81 and a non-person image window 82. The person image window 81 displays the discrimination target images I1, I2, I4-I6 that were discriminated as person images before the reference value setting was changed. The non-person image window 82 displays the discrimination target images I3 and I7 that were discriminated as non-person images before the reference value setting was changed. The reference value setting is changed, and discrimination processing is performed again, so that the discrimination target image I4 that was discriminated as a person image is now discriminated as a non-person image. Then, a frame FL is attached to the discrimination target image I4 by the CPU 2 (an example of a notification control means). The user can see that the discrimination result has changed due to the change in the reference value because the frame FL has been attached (an example of a change in the notification mode). The discrimination target image I4 was discriminated as a person image before the reference value was changed, but has been discriminated as a non-person image due to the change in the reference value. 23, before the reference value is changed, the discrimination target image I4 is displayed in a person image window 81 (one example of a notification means), and even when the reference value is changed and the discrimination result changes from a person image to a non-person image, it is still displayed in the person image window 81, but as a result of the change in the reference value, the discrimination target image I4 may be switched from display in the person image window 81 to display in the non-person image window 82 (one example of changing the notification mode). In that case, a frame FL may be added so that it is easy to understand that the discrimination result has changed, but it does not have to be added.

[0106] Returning to FIG. 22, if the reference value setting is not changed (NO in step 171), the discrimination result does not change (NO in step 174), or a frame is added to the image to be discriminated according to the discrimination result (step 175), image data representing the next image to be discriminated that has not been subjected to discrimination processing is read (step 34 in FIG. 21), and discrimination processing is performed (step 35).

[0107] FIG. 24 shows a modified example, and shows a part of a flowchart showing the processing procedure of the image discrimination device 1.

[0108] The process shown in FIG. 24 is carried out following the process of step 36 shown in FIG.

[0109] When the discrimination process is completed for all discrimination target images (NO in step 37), the CPU 2 (adoption decision means) decides whether or not to use the discrimination target images discriminated as human images and the discrimination target images discriminated as non-human images in the album (step 181). For example, when creating an album using only human images without using non-human images, a human image to be used in the album is decided from the images discriminated as human images based on the degree of blur, the degree of brightness, the position of the main person, etc., and the paste position of the decided human image is decided by the CPU 2 (position decision means) to create the album (step 182). The paste position may be decided by the user. Conversely, when creating an album using only non-human images without using human images, a non-human image to be used in the album is decided from the images discriminated as non-human images (step 181), and the paste position of the decided non-human image is decided by the CPU 2 (position decision means) to create the album (step 182). In this way, for example, an album of only images discriminated as human images and an album of only images discriminated as non-human images can be created.

[0110] Also, instead of an album of only images determined to be human images or an album of only images determined to be non-human images, an album may be created from images determined to be human images and images determined to be non-human images. For example, images to be used in the album may be determined by the CPU 2 (use determining means) so that the ratio between images determined to be human images and images determined to be non-human images is a predetermined ratio (step 181), and the CPU 2 (position determining means) may determine the paste position of the determined images to create an album (step 182).

[0111] According to the third embodiment, it is possible to know that the discrimination result of the discrimination target image has been changed due to the change in the reference value. Also, it becomes possible to create an album according to the discrimination result of the discrimination target image.

[0112] In the above embodiment, many images to be discriminated were discriminated into human images and non-human images, but after discriminating them as human images (first stage discrimination processing), the human images may be further discriminated into "human images that are more human-like" and "human images that are closer to non-human" (second stage discrimination processing). For example, for a certain criterion item, a sub-slider may be provided at a position closer to the direction of greater human degree than the above-mentioned slider, and the human degree indicated by this sub-slider may be used as a reference value to discriminate between "human images that are more human-like" and "human images that are closer to non-human". Similarly, non-human images may be further discriminated into "non-human images that are more non-human-like" and "non-human images that are closer to humans".

[0113] In the above-described embodiment, the image discrimination device 1 may be realized using a personal computer or the like, or may be realized using a multi-function mobile phone such as a smartphone, or a flat-panel computer such as a tablet.

[0114] The processing units that execute the above-mentioned processes include the CPU 2, which executes software and functions as various processing units, as well as programmable logic devices such as FPGAs (field-programmable gate arrays) whose circuit configuration can be changed after manufacture, and dedicated electrical circuits such as ASICs (application specific integrated circuits), which are processors with a circuit configuration specially designed to execute specific processes.

[0115] One processing unit may be configured with one of these various processors, or may be configured with a combination of two or more processors of the same or different types (for example, multiple FPGAs, or a combination of a CPU and an FPGA). As an example of configuring multiple processing units with one processor, first, as represented by computers such as client computers and servers, one processor is configured with a combination of one or more CPUs and software, and this processor functions as multiple processing units. Second, as represented by systems on chips, there is a form in which a processor is used that realizes the functions of the entire system including multiple processing units with a single IC (integrated circuit) chip. In this way, the various processing units are configured using one or more processors of various types as a hardware structure.

[0116] Furthermore, the hardware structure of these various processors is, more specifically, an electrical circuit that combines circuit elements such as semiconductor elements. [Explanation of symbols]

[0117] 1: Image discrimination device, 2: CPU, 3: Display device, 4: Communication device, 5: Hard disk, 6: Hard disk drive, 7: Memory, 8: Keyboard, 9: Mouse, 10: Compact disk drive, 11: Compact disk, 12: Memory card reader / writer, 13: Memory card, 20: Artificial intelligence device, 21: Database, 22: Feature extraction device, 23: Learning algorithm, 24: Learning model, 25: Communication device, 40: Image display window, 41: Image display area, 42: Slide bar, 43: Knob, 50: Reference value setting window, 51: First check box, 52: First slide bar, 53: First slider, 61: second check box, 62: second slide bar, 63: second slider, 71: second check box, 72: third slide bar, 73: third slider, 80: image classification window, 81: person image window, 82: non-person image window, 90: facing page, 91-93: image insertion area, 121: non-person part, 122: person part, 131-132: image, 150: classification result display window, 151-154: text, FL: frame, I: sample image, I1-I28: classification target image, I31-I33: classification target image, I41-I43: classification target image, I51-I53: classification target image, L1-L3: distance

Claims

1. an evaluation value input means for inputting an evaluation value of the likeliness of the discrimination target image as a human image; a first display control means for controlling to display a reference value setting image on which a plurality of reference items used for determining the likelihood of a human image for determining whether the discrimination target image is a human image are displayed; a criterion item changing means for changing the criterion item used for determining whether or not the image is a human image based on whether or not a user has selected the criterion item, among the plurality of criterion items displayed in the reference value setting image displayed under the control of the first display control means; a criterion value input means for receiving an input of a criterion value related to the criterion item changed by the criterion item changing means from a user; a reference value setting means for setting the reference value received by the reference value input means as a reference value to be used in a process of discriminating a human image; a discrimination means for discriminating the discrimination target image as a human image when an evaluation value of the discrimination target image inputted from the evaluation value input means is equal to or greater than the reference value; a notification means for notifying a result of the discrimination means; and a notification control means for changing a notification mode of the discrimination result notified by the notification means from a notification mode before the discrimination result is changed in response to a change in the discrimination result in the discrimination means due to a change in the reference value set by the reference value setting means; An image discrimination device comprising:

2. The discrimination means is a learning model that has been trained to have a function of outputting an evaluation value of the likeliness of an input image using a plurality of sample images to which evaluation values ​​of likeliness of a human image have been assigned, and the learning model determines that the discrimination target image is a human image when the evaluation value of the likeliness of the discrimination target image obtained by inputting the discrimination target image is equal to or greater than the reference value; The image discrimination device according to claim 1 .

3. The above criteria are: a first area ratio of an area of ​​a part of a person included in the discrimination target image to an area of ​​the entire discrimination target image; and The discrimination means is When the first area ratio is equal to or greater than the reference value, the discrimination target image is discriminated as a human image. The image discrimination device according to claim 1 .

4. a second display control means for displaying the discrimination target image by changing the display manner of the person portion in the discrimination target image from the display manner of the other portions; The image discrimination device according to claim 3 , further comprising:

5. The above criteria are: a second area ratio of an area of ​​a portion of a person included in the discrimination target image to an area of ​​a portion of an object other than the person included in the discrimination target image; The discrimination means is When the second area ratio is equal to or greater than the reference value, the discrimination target image is discriminated as a human image. The image discrimination device according to claim 1 .

6. The above criteria are: a third area ratio of an area of ​​a person portion included in the discrimination target image to an area of ​​a landscape portion included in the discrimination target image, The discrimination means is When the third area ratio is equal to or greater than the reference value, the discrimination target image is discriminated as a human image. The image discrimination device according to claim 1 .

7. an adoption or non-adoption determining means for determining, when the discrimination result in the discrimination means is discriminated as a portrait image, that the discrimination target image judged to be a portrait image should be adopted in an album using only portrait images; The image discrimination device according to claim 1 , further comprising:

8. The evaluation value input means inputs an evaluation value of the likelihood that the image to be discriminated is a human image, a display control means for controlling display of a reference value setting image in which a plurality of reference items used for determining whether the discrimination target image is a human image or not are displayed; a criterion item changing means for changing the criterion item used for determining whether or not the image is a human image based on whether or not a user has selected the criterion item, among the plurality of criterion items displayed in the reference value setting image displayed under the control of the first display control means; a criterion value input means for receiving, from a user, an input of a criterion value related to the criterion item changed by the criterion item changing means; a reference value setting means for setting the reference value received by the reference value input means as a reference value to be used in a process of discriminating a person image; a discrimination means for discriminating the discrimination target image as a human image when the evaluation value of the discrimination target image inputted from the evaluation value input means is equal to or greater than the reference value; The notification means notifies the result of the determination by the determination means, a notification control means for changing a mode of notification of the discrimination result notified by the notification means from a mode of notification before the discrimination result is changed in response to a change in the discrimination result in the discrimination means due to a change in the setting of the reference value in the reference value setting means; Image discrimination method.

9. A computer-readable program for controlling a computer of an image discrimination device, The user is asked to input an evaluation value of the likelihood that the target image is a human image, displaying a reference value setting image in which a plurality of reference items are displayed to be used for determining whether the discrimination target image is a human image or not; changing the criterion item used for determining whether or not the image is a human image based on whether or not a user has selected the criterion item among the plurality of criterion items displayed in the reference value setting image to be displayed; Accepting input of the changed criterion values ​​for the criterion items from the user; The accepted reference value is set as a reference value to be used in a process of discriminating a person image; When the evaluation value of the likeliness of the input discrimination target image as a human image is equal to or greater than the reference value, the discrimination target image is discriminated as a human image, The discrimination result is notified, A program for controlling a computer of an image discrimination device so as to change the manner in which the discrimination result is notified from the manner in which it was notified before the discrimination result was changed in response to a change in the setting of the reference value.

10. A recording medium storing the program according to claim 9.

11. There are multiple criteria above, The discrimination means is For each criterion item selected by the user, an individual discrimination value indicating the difference between the value of the criterion item of the discrimination target image and a reference value is calculated, and a comprehensive discrimination value is calculated by adding up the calculated individual discrimination values; If the overall discrimination value is equal to or greater than a threshold value, the image to be discriminated is discriminated as a human image. The image discrimination device according to claim 1 .

12. The first display control means Display a slider to set the reference value, The control unit further includes a reference value changing unit for changing the reference value by a user operating the slider, The reference value input means is The value changed by the reference value changing means is accepted as the reference value. The image discrimination device according to claim 1 .

Citation Information

Patent Citations

  • Rolling device of reaper in combined harvester

    JP1981055112A

  • Image printer and printing method

    JP2002187329A

  • Discriminator setting device, degree-of-attention measuring device, discriminator setting method, degree-of-attention measuring method, and program

    JP2006065447A

  • Method, apparatus and program for image identification, image processing method and integrated circuit

    JP2006268820A

  • Image selection method, image selection device, program and print order accepting machine

    JP2006318034A