JP2010262531A - Image information processor, image information processing method and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To solve the following problem: when selecting an excellent image from images obtained by photography, an image list wherein the images are merely arranged in photographic order is inconvenient because the target image cannot be selected unless all the images are viewed. <P>SOLUTION: This image information processor learns last image selection, and makes it be applied to next image selection of a new image. The image information processor stores a selection reference value of selection/non-selection and an effective classification method by a difference in a characteristic amount between a selection image and a non-selection image inside an image group of a classification candidate by various kinds of classification methods. The image information processor classifies a new image group by the learned classification method, and determines order of the images by the stored selection reference value, so that the image to be selected is easy to be selected after the learning. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention makes it easy to select an image desired by a user from a plurality of images in accordance with a user operation instruction.

  Conventionally, when a desired image is found from a plurality of captured images, the desired image is searched by looking at all the images one by one while searching from the image list display or comparing each image.

  In order to make it easy to find a desired image, the automatic image classification method disclosed in Patent Document 1 prepares an image for each category having different image characteristics, and based on the feature points of the image, the image is classified from a large category. It was the structure classified hierarchically into the detailed classification category registered in the category of large classification.

  The digital still camera disclosed in Patent Document 2 is configured by a recording medium that records detected camera shake amount data in association with a corresponding subject image. Then, the subject images are sorted in descending order of camera shake amount, and the reproduced images of the subject images are displayed together with the camera shake amount data in descending order of camera shake amount.

  In the image selection method disclosed in Patent Literature 3, a plurality of candidate images are stored, and the user model storage unit stores a user model suitable for the user's preference. For the candidate images, the strength of the induction field, the potential Construct line complexity, induction field energy and image. The configuration is such that the luminance values of the three primary colors of each pixel are extracted as image feature information, and an image that seems to fit the user's preference is displayed based on the extracted image feature information and the user model.

JP 2005-309535 A JP 2003-134389 A JP 2003-256467 A

  However, the automatic image classification method disclosed in Patent Document 1 targets images of a predetermined classification. Therefore, processing cannot be performed unless the user designates the image classification. Alternatively, it is not considered that the image classification changes with time. If the user-specified classification is ambiguous, images that should originally be specified in different classifications are set to the same classification. The user needed to specify the correct classification.

  In the digital camera disclosed in Patent Document 2, the camera shake amounts are sorted according to the camera shake amount, so that it is easy to select an image without camera shake. However, camera shake may not be a selection criterion for some images. In addition, it is impossible with this configuration to adapt to the case of a plurality of standards or a standard changed by the user.

  The image selection method disclosed in Patent Document 3 has a configuration in which user preferences can be modeled. There are few cases where image preferences can be modeled by values such as image characteristics (luminance values of the three primary colors of pixels). For example, when an image of a user's child is to be used for an album, an example in which the priority of an image showing another child's child is low can be considered. In the above example, it is unlikely to be possible to extract image preferences from image features.

  Therefore, an object of the present invention is to enable automatic learning even when a user selects an image based on a different standard for each photographing target.

  An image information processing apparatus of the present invention includes an image selection instruction unit that instructs an arbitrary image, an image classification unit that classifies a selected image based on a classification condition, an extraction unit that extracts a selection criterion for each classification, And means for storing the classification conditions and selection criteria, and means for adapting image processing based on the stored classification conditions and selection criteria.

  According to the present invention, even if the user selects an image based on a different standard for each shooting target, the user is automatically learned. Since a desired image is always displayed at the top of the display list after learning, it is easy to select. In an image processing apparatus in which images are automatically selected and arranged, a desired image is automatically arranged if the present invention is incorporated.

It is a figure showing the actual usage pattern in the 1st Embodiment of this invention. It is a hardware block diagram of the image information processing apparatus which concerns on the 1st Embodiment of this invention. It is a software block diagram of the image information processing apparatus which concerns on the 1st Embodiment of this invention. It is a figure which shows the example of a general image selection display screen. It is a classification conceptual diagram of an image group. It is a flowchart showing the process which finds a selection reference | standard by the some classification method. It is a flowchart which shows the detail of the selection reference value extraction process for every classification | category which classified the image with imaging | photography time. It is a flowchart which shows the process which classify | categorizes an image by image similarity and extracts selection criteria from the classified image. It is a flowchart which shows the selection reference | standard extraction process of the classification image in an imaging | photography place. It is a figure which shows the structure of a learning dictionary. It is a flowchart which shows a learning result adaptation process. It is a figure which shows the example of a display before performing after performing the image processing based on learning data. It is a software block diagram of the image information processing apparatus which concerns on 2nd Embodiment of this invention. It is a conceptual diagram showing the structure of the learning dictionary in the 2nd Embodiment of this invention. It is a flowchart showing a learning adaptation process and a learning process. It is a flowchart showing a learning adaptation process and a learning process. It is a figure which shows the state of an image list display.

  DESCRIPTION OF EXEMPLARY EMBODIMENTS Hereinafter, preferred embodiments to which the invention is applied will be described in detail with reference to the accompanying drawings.

<First Embodiment>
First, a first embodiment of the present invention will be described. FIG. 1 shows an actual usage pattern in the first embodiment of the present invention. Reference numeral 1 denotes an image information processing apparatus according to the present embodiment, which includes general components. Software for displaying still images recorded on a recording medium and processing procedures according to the present embodiment are incorporated. 2 is a pen for a general transparent resistance film digitizer.

  FIG. 2 is a hardware configuration diagram of the image information processing apparatus according to the present embodiment. 6 is a position coordinate input unit composed of a general transparent resistance film digitizer. When the surface is pressed down with a pen or the like, the XY position coordinate data of the pressed position is 8 CPUs via 12 system buses. Sent to. Reference numeral 7 denotes a general liquid crystal display unit, which includes a liquid crystal display element, a liquid crystal control circuit, and a display memory, and is connected to 8 CPUs via 12 system buses. In response to an instruction from the CPU 8, image display and character display are executed. Reference numeral 8 denotes a general CPU, which is connected to a RAM 9, a ROM 10, a liquid crystal display unit 7, and the like via a system bus 12, and performs processing operations by a program stored in the ROM 10. Do. Reference numeral 9 denotes a general RAM, which is used as a work area. Reference numeral 10 denotes a general ROM in which the image processing procedure of this patent is stored. Reference numeral 11 denotes an SD card which is a memory card used in general information equipment, and image information and the like are stored in this equipment connected to 7 CPUs via 12 system buses. A general system bus 12 exchanges data between the CPU 8 and the ROM 10, the RAM 9, and other devices.

  FIG. 3 is a software configuration diagram of the image information processing apparatus according to the present embodiment. An image storage means 13 reads and stores an image stored in the memory card 11 of FIG. The image is a file format such as a general JPEG. Reference numeral 14 denotes an image means for performing image classification according to a rule in which images are incorporated in advance. Reference numeral 15 denotes a selection criterion extraction unit for each classification, which extracts selection criteria for classified (selected & non-selected) images. Extract feature values of various built-in images such as differences in metadata between selected images and non-selected images of classified images, image brightness, image contrast values, image focus determination values, etc. Features of features that are clearly different in non-selected images are extracted as selection criteria. Reference numeral 16 denotes learning storage means for storing the image classification conditions and image selection criteria determined in 14 and 15. A learning selection priority adaptation unit 17 determines an image to be displayed at present from the classification conditions of the 16 learning storage unit and an image selection criterion, and performs control so that an image to be selected by the user is displayed preferentially.

  FIG. 4 is a diagram illustrating an example of a general image selection display screen. Reference numeral 19 in FIG. 4 denotes an image display area. This is an area for displaying a reduced image. Reference numeral 18 in FIG. 4 denotes a general scroll bar area. By operating the black part, the image in 19 in FIG. 4 is scrolled in conjunction with the operation. When the pen shown in FIG. 1 is pressed on the image 22 displayed in FIG. 4, it is temporarily selected. Reference numeral 20 in FIG. 4 denotes an image selection determination button. When this button is pressed, the image temporarily selected at that time is selected. Reference numeral 21 in FIG. 4 denotes a cancel button that puts the selected image into a non-selected state. Reference numeral 22 in FIG. 4 is a display example of the selected image. The selected character is displayed on the selected image. The image selection instruction means is realized by the above processing.

  FIG. 5 is a conceptual diagram of image group classification. This indicates that there are different selection criteria for each image classification. One circle represents one type of image group. The straight line dividing the circle represents the selection criterion. For example, a group of 23 images A is a travel image classification. The classification conditions are classified based on the shooting date based on the date when the trip was made. In the case of a user, the selection criteria for the images in the group A are those with good composition and appropriate contrast. The selected image and the non-selected image are divided according to the selection criteria. The group of 24 images B is a child image classification. The classification criterion is an image of your child. It is classified according to the degree of coincidence with the registered image, etc. by the image entered with the child in the metadata of the image file, or the similar image classification of the image. As an example of a selection criterion in the classification of the child images, an image in which the child is greatly reflected is given priority. When shooting continuously in the same scene, the one that is in focus is given priority. In addition, the 25 image groups C are hobby images. In this example, the classification standard is information on latitude and longitude added to the image. As a selection criterion in this hobby image classification, priority is given to a specific subject being shown large, and priority is given to an object with appropriate focus and contrast.

  As described above, when various images are taken, stored, and selected when creating an album or the like, there are many classification criteria and selection criteria for each image classification. In this example, the image A group and the image B group are separated from each other, but in some cases, a part thereof may cross over. For example, if an image of a child is taken on a trip, the image is classified into both groups A and B.

  FIG. 10 is a conceptual diagram of the structure of the learning dictionary. The learning dictionary is configured as in the example of FIG. One dictionary data stores a dictionary number, a classification condition type, a classification condition value, a selection criterion type, and a selection criterion value.

  FIG. 12 shows display examples before and after image processing based on learning data. The image display example 26 in FIG. 12 is an example of an image list before the processing of this embodiment is performed, and the images are arranged, for example, from the upper left to the lower right in the order of photographing dates. There are 27, 28, 29, and 30 images, and it is assumed that 27 are images selected by the user. When the selection criterion is extracted by this processing and the selection image is displayed with priority from the selection criterion, the selection criterion is rearranged as 31 in FIG. Non-selected candidate images such as 28 and 29 move to the rear of the list.

  Processing will be described with reference to the flowcharts of FIGS. 6, 7, 8, 9, and 11. The flowchart of FIG. 6 is a flowchart showing processing for finding selection criteria by a plurality of classification methods. This corresponds to the image classification means 14 in FIG. 3 and the selection criterion extraction means 15 for each classification in FIG.

  It is assumed that there are a plurality of classifications in the image and the selection criteria are different for each classification. Depending on the user, there may be a case where only a specific image is taken. In such a case, there is no problem with one selection criterion. However, there are many cases in which shooting is usually performed with different criteria for each subject or event. .

  In step S6-1, a process of finding selection criteria by a plurality of classification methods is started. Initialize the work area. Load an image file into the work area.

  In step S6-2, a process of classifying the images by the shooting time is performed. For example, it seems that there are many examples of shooting different subjects under different shooting conditions at day and night, and therefore, for example, morning, noon, and night are classified according to the shooting time. In this embodiment, processing in this shooting time classification will be described. Read the shooting date and time information of the target image. For example, AM 4 o'clock to 9 o'clock is classified as morning, AM 9 o'clock 1 minute to PM 5 o'clock, and the others are classified as night. An example where the time itself is determined within a range that takes into account the season and latitude and longitude is also conceivable. An embodiment in which shooting conditions such as sunshine conditions are taken into account depending on the time and region is also conceivable. In addition, if a seasonal event is to be taken, a method of classifying the shooting date by month may be considered.

  In step S6-3, a process for extracting a selection criterion is performed by examining the difference between each feature of the selected image and the non-selected image for each classification. Details of this processing will be described with reference to the flowchart of FIG.

  In step S6-4, it is checked whether a feature is not found in step S6-3. If no feature is found, the process proceeds to step S6-6, and if found, the process proceeds to step S6-5. If a feature is found in each of the morning, noon, and night categories, the category and features are stored in the learning dictionary. If only the morning classification is found in the selection criteria, it is recorded in the learning dictionary. If the feature of the selection criterion cannot be extracted for all morning, noon, and night classifications, the process proceeds to step S6-6 of the next different classification process.

  In step S6-5, a process of storing the found classification method and selection criterion features in the learning dictionary is performed. For example, if the classification condition of the condition found in step S6-3 is for an image with a shooting time of 6 to 8 o'clock, that is stored. If the selection condition is contrast and the range of the contrast value is A to B, it is stored.

  In step S6-6, classification is performed based on the feature amount of the image, and a process of extracting a selection criterion is performed by examining a difference between each feature of the selected image and the non-selected image for each classification. Details of the classification will be described with reference to the flowchart of FIG.

  In step S6-8, it is checked whether a selection criterion has been found. If the selection criterion is found, the process proceeds to step S6-5, and the selection criterion is stored in the learning dictionary. If no selection criterion is found, the process proceeds to step S6-9.

  In step S6-9, a process of classifying according to the shooting location latitude / longitude information of the image metadata is executed.

  Sort images by latitude and longitude information of shooting location in image metadata. Objects in the vicinity of longitude and latitude are in the same group. The latitude / longitude information is obtained from Exif information, which is image metadata of all images, and stored. The latitude / longitude information of the Exif information may be entered with the camera or added later. Thereafter, the images are sorted according to the stored latitude / longitude information, and images having the same latitude / longitude and neighboring latitude / longitude are classified as one group. Numbers below the latitude and longitude frequencies are treated as the same. The image of the same city is put together by the latitude and longitude information. For example, if the number of objects to be sorted is 100, the group is classified so that one group becomes 10 or more within the same latitude and longitude ± α.

  In step S6-10, a selection criterion is extracted by selecting each category classified by the shooting location latitude / longitude information and examining a feature difference between non-selected images. Details will be described with reference to the flowchart of FIG.

  In step S6-11, it is checked whether a selection criterion has been found in the shooting location classification. If a selection criterion is found, the process proceeds to step 6-5, and if not, the process proceeds to step S6-12.

  In step S6-12, the flowchart process of FIG. 6 ends. If no feature of the selection criterion is found in all classifications, the learning dictionary is not stored, but the selection criterion is usually detected and stored in the learning dictionary in some classification.

  The details of the selection reference value extraction processing for each classification in which the images are classified by the photographing time in the flowchart of FIG. 7 will be described.

  In step S7-1, a process for obtaining a selection reference value for each classified image classified by the photographing time is started. In the information of the shooting time of each image, the image from AM 4 to 9 is classified as morning, AM 9:01 to PM 5 is classified as daytime, and the others are classified as night. The images are classified into a morning image group, a daytime image group, and a night image group.

  In step S7-2, processing for reading an image of one category is performed. First, images classified in the morning are read into the work area.

  In step S7-3, a process for obtaining the difference between the feature quantity 1 between the selected image and the non-selected image is performed. A feature amount 1 is a feature amount of an image edge in the focus area. Based on this feature amount, it is evaluated whether the image selection criterion is that the image in the focus area is in focus. If the difference between the average value of the feature amount 1 of the selected image and the average value of the feature amount 1 of the non-selected image is equal to or greater than a specified value, it is determined that the selection criterion is used. In this step, the difference between the average value of the feature value 1 of the selected image and the average value of the feature value 1 of the non-selected image is stored.

  In step S7-4, a process for obtaining a difference between the feature amount 2 between the selected image and the non-selected image is performed. The feature amount 2 is the brightness of the image. The difference between the average brightness value of the selected image and the average brightness value of the non-selected image is stored as the difference of the feature amount 2.

  In step S7-5, a process for obtaining a difference between the feature amount 3 between the selected image and the non-selected image is performed. The feature amount 3 is the image level. The level of the image is calculated based on the building or horizon in the image. In the case of a landscape image or the like, it seems that it is better to be horizontal rather than obliquely displayed, so this process is used to check whether it is a selection criterion. The difference between the average horizontal value of the selected images and the average horizontal value of the non-selected images is stored.

  In step S7-6, it is checked whether the difference between the feature amounts is equal to or greater than a threshold value. If the difference between all the feature amounts is equal to or smaller than the threshold value, the process proceeds to step S7-8 to check whether the processing of all the classified images is completed. Perform processing.

  In step S7-7, processing for storing feature quantities and classification conditions equal to or greater than a threshold value is performed. If the feature quantities 1, 2, and 3 are greater than or equal to the threshold value, a process for storing the value and the classification time of the classification condition in the learning dictionary is performed.

  In step S7-8, it is checked whether all classifications have been processed. If the morning, noon, and night classifications have been processed, the process proceeds to step S7-9. If not, the process proceeds to step S7-2 to perform processing for reading the next classification image.

  In step S7-9, the process of this subroutine is terminated and the process proceeds to the next classification method.

  In the flowchart processing of FIG. 7, three feature amounts are extracted in order to simplify the description. However, in order to deal with various users, it is necessary to check other feature amounts, and it is easy to add other than these three. In addition, there are cases where the user adopts other three feature quantities as well as the three adopted ones. The type of the feature amount may be changed at the time of setting.

  A process of classifying images by image similarity and extracting selection criteria from the classified images will be described with reference to the flowchart of FIG. In many cases, the image of the subject you are paying attention to is important for your child's image, and the surrounding landscape is not important. In such a case, it is considered that it is easier to extract selection criteria for selection and non-selection if classification is based on images in the focus area. Such an image is processed by this classification method.

  Or, for example, in the case of a mountain image, an underwater image, a night view image, and different types of scenes, if the image selection criteria are different, a method of classifying by the RGB values of the entire image may be considered.

  In step S <b> 8-1, a process of classifying the images by the similarity of the images and extracting a selection criterion for each classified image is started.

  In step S8-2, processing for reading one image file into the work area is performed. In step S8-3, information on the focus area is read from the image metadata, and an image around the designated focus area is read into the work area.

  In step S8-4, processing is performed to store an image having a wide area where the edge between pixels is clear from the read image as an image object. Alternatively, in the case of an image intended for a person, a face object can be stored using an algorithm such as face recognition. A conventional algorithm may be used as the face recognition algorithm itself.

  In step S8-5, the image features of the object stored in step S8-4 are sorted and stored. For example, image features differ for a child's face, a cat's face, and a car. The child's face and cat's face are classified by the same feature amount.

  In step S8-6, it is checked whether all the images have been processed. If completed, the process proceeds to step S8-7, and if not, the process proceeds to step S8-2.

In step S8-7, the difference between the feature amount of the selected image and the feature amount of the non-selected image is checked for each feature amount. If your child's face classification is in focus and a laughing image is the selection criterion, the value of the smiling face in the selected image and the non-selected image will differ depending on whether the image is in focus or not. Get that value. If the value is equal to or greater than the threshold value, the process proceeds to step S8-8. Otherwise, in step S8-8, the image, the image feature amount, and Exif information that is image metadata are stored in the learning dictionary. For example, in the face classification of one's own child, a face image, a value obtained by quantifying the way the focus is selected, and a value obtained by quantifying a smile are stored. For example, the conventional algorithm described in Japanese Patent Application Laid-Open No. 2004-46591 may be used to digitize the smile.
In step S8-9, this process is terminated, and the process proceeds to the next process.

  FIG. 9 is a flowchart showing selection criteria extraction processing of the classified image at the shooting location. In step S9-1, a process of extracting selection criteria from each classification at the shooting location is started. Secure and initialize the work area.

  In step S9-2, the image of one group classified in the step is read into the work area.

  In step S9-3, Exif information that is additional information of the selected image and the non-selected image is compared. For example, in a selected image of a certain user, when an image taken with manual focus is special and an image taken with autofocus is not selected, information on manual focus is extracted.

  In step S9-4, it is checked whether there is Exif information having a difference. If there is, the process proceeds to step S9-6, and if not, the process proceeds to step S9-5.

  In step S9-6, the classification condition and Exif information which is additional information of the image are stored in the learning dictionary.

  In step S9-5, for images that have been continuously shot, or for images that have approximately the same shooting time, the image similarity of the entire image is measured, and if it is less than or equal to the threshold value, it is stored as an image of the same scene.

  In step S9-7, the focus around the focus area of the selected image and the non-selected image in the similar images classified in step S9-5 is checked. By checking the difference in the edge of the image pixel, it is checked whether the image is in focus or blurred. A conventional algorithm may be used.

  In step S9-8, if the selected image is in focus and the non-selected image is not in focus, the process proceeds to step S9-9, and if the focus is not a condition, the process proceeds to step S9-10.

  In step S9-9, the latitude / longitude information of the classification condition and the focus condition of the image in focus are stored in the learning dictionary.

  In step S9-10, it is checked whether all groups have been processed. If all groups have been processed, the process proceeds to step S9-11, and if not, the process proceeds to step S9-2.

  In step S9-11, the selection criterion extraction process in the shooting location classification is terminated. In this way, various classification methods are tried, and by checking the difference between selected and non-selected images for each classification, a selection criterion for a certain classification is obtained and stored in the learning dictionary shown in FIG. Go.

  In order to adapt to a general user, it is better to create an initial learning dictionary in advance. For example, ask multiple general users to take images and create albums from the images. The initial learning dictionary is created by providing the selected and unselected image data in the process. Then, when it is used for the first time, it is possible to realize a general priority image presentation in the initial learning dictionary.

  FIG. 11 is a flowchart showing learning result adaptation processing. In step S11-1, a learning adaptation process is started. Secure and initialize the work area. This processing is started when the memory card of FIG. 2 in which a new image is stored is inserted while the learning dictionary of FIG. 10 is created.

  In step S11-2, a process of reading one image in the memory card 11 in FIG. By sequentially executing this step, the image in the memory card is read into the work area.

  In step S11-3, a process of reading one learning data from the learning dictionary of FIG. 10 is performed. Data are sequentially read from the data of dictionary number 1 in FIG.

  In step S11-4, it is checked whether the image read in step S11-2 matches the dictionary classification conditions. If the classification condition is met, the process proceeds to step S11-5, and if not, the process proceeds to step S11-7.

  In step S11-5, it is checked whether the selection condition is met. If the selection condition is met, the process proceeds to step S11-6, and if not, the process proceeds to step S11-7.

  In step S11-6, an image that matches the selection condition is stored. The image is stored in the selection condition storage buffer, and when the image list is displayed, the image is displayed at the beginning as a selection priority image to facilitate selection by the user.

  In step S11-7, it is checked whether all learning dictionary data have been checked. If all the learning dictionaries are checked, the process proceeds to step S11-8, and if there is no check, the process proceeds to step S11-3.

  In step S11-8, it is checked whether all image data has been checked. If all the images have been processed, the process proceeds to step S11-9. If there is an unprocessed image, the process proceeds to step S11-2.

  In step S11-9, the stored selected image is output. When processed in this way, as shown in FIG. 12, the images in the 26 shooting orders are sorted and displayed as images to be selected. For example, non-selected candidate images such as 28, 29, and 30 go to the back of the list. This process ends in step S11-10. In this embodiment, an example in which an image that matches the selection criteria is displayed at the top of the list display in the image selection list display has been described. However, an example in which the present invention is implemented in an image application that selects an image and creates an image album is also considered. It is done. In that case, an image that matches the selection criteria of the learning data can be used as a selected image for processing to automatically arrange in the image arrangement area of the image album. The image automatic arrangement processing is realized by the above processing.

  By configuring the processing in this way, it is possible to realize an apparatus that preferentially displays an image that should be selected by the user based on the user's image selection criteria.

<Second Embodiment>
Next, a second embodiment of the present invention will be described. In the first embodiment, the image to be selected is selected when the target image matches the dictionary data condition of the learning dictionary. In this configuration, learning dictionary data created with a small number of image data in the past is also effective, so in some cases, the selection criterion may be inappropriate for a learning dictionary created with a small number of image data. A configuration in which dictionary data is validated after collecting a certain amount of time or sample data is conceivable. In the second embodiment, learning adaptation determination means and image enhancement display means are provided in addition to the configuration of the first embodiment.

  FIG. 13 is a software configuration diagram of the image information processing apparatus according to the second embodiment of the present invention. The same components as those in the first embodiment are denoted by the same part numbers and will not be described. Reference numeral 32 in FIG. 13 denotes learning history determination means. A process of determining whether to apply the learning data from the history of the dictionary data of the learning dictionary is performed. Reference numeral 33 denotes an image emphasizing display means for emphasizing so that the user can easily select an image to be selected.

  FIG. 14 is a conceptual diagram showing the structure of the learning dictionary in the second embodiment of the present invention. One dictionary data includes a dictionary number, a classification condition type, a classification condition value, a selection condition type, a selection condition value, and a learning target image amount / adaptive history / time. In the process of extracting the selection criterion from the selected non-selected images, the number of target images is stored as the learning target image amount. This processing is learning data creation target image amount measurement processing.

  15 and 16 are flowcharts showing the learning adaptation process and the learning process. Processing will be described with reference to this flowchart. FIG. 15 is a flowchart showing the learning adaptation process. The process starts with the insertion of the memory card 11 in FIG. 2 in which the learning dictionary shown in FIG. 14 exists and the target image is stored.

  In step S15-1, a learning adaptation process is started. The work area is secured and initialized, and the dictionary file is read. In step S15-2, one image data is read into the work area from the memory card 11 in FIG.

  In step S15-3, one learning data in the learning dictionary in FIG. 14 is read into the work area. Read from dictionary number 1.

  In step S15-4, it is checked whether the image data matches the classification condition of the learning data. If they match, the process proceeds to step S15-5, and if they do not match, the process proceeds to step S15-8.

  In the example of the dictionary data 1 in FIG. 14, the image data matches when the shooting time of the image data is from 6:00 to 8:00, and does not match otherwise.

  In step S15-5, it is checked whether the image data matches the learning data selection condition. If the selection condition is met, the process proceeds to step S15-6, and if not, the process proceeds to step S15-8. In the example of the dictionary data 1 in FIG. 14, it matches if the contrast value of the image data is between A and B, and does not match otherwise.

  In step S15-6, the learning amount & learning time is checked. Here, the learning state of the dictionary data of the dictionary is checked to determine whether to adapt. If the prescribed creation target image amount is satisfied, the process proceeds to step S15-7, and if it is less than the prescribed creation target image amount, the process proceeds to step S15-8. For example, dictionary data created in a state where three images are selected from ten images may be an accidental example, and is inappropriate as learning data. For example, dictionary data created from 100 images is trusted and adapted. Or, since it may be unstable if it is a short-term study, check the study time.

  In step S15-7, a selection mark is added to the image file, and a process for adding the learning adaptation history is performed. Add how much you adapt. Make images that tend to be used more easily.

  In step S15-8, it is checked whether all learning data has been checked. If all the learning data are checked, the process proceeds to step S15-9, and if not, the process proceeds to step S15-3.

  In step S15-9, it is checked whether all image data has been checked. If all the image data has been processed, the process proceeds to step S15-10, and if there is unprocessed image data, the process proceeds to step S15-2.

  In step S15-10, the stored selected image is output. The output result of the selected image is as indicated by 34 in FIG. For example, a good image to be selected by the user may be highlighted as shown, and an image that is not normally selected may be displayed as small as displayed. As in the example, if you select a large image to be selected, it will be easy to select. Alternatively, it may be possible to blink the frame of the image. Alternatively, a method may be considered in which the image brightness of the non-selected candidate image is lowered and the image having a strong tendency to select is displayed more vividly.

  In step S15-11, this process ends. Discard the work area and close the dictionary file.

  The learning process will be described along the flowchart of FIG. Since there is a case where an image is selected from among the preferentially displayed images or other images are selected, the learning process in the instruction operation will be described.

  In step S16-1, a learning process is started. This process is called in a state where an image list display as shown in FIG. 17 is displayed.

  In step S16-2, a learning-adapted image list is displayed. The display example is as shown in FIG.

  In step S16-3, it is checked whether the preferentially displayed image is selected as the album image. When selected as an album image, check if a selection mark is present. For example, if 32 images in FIG. 17 are selected and instructed as album images, the image 35 in FIG. 17 is highlighted and the process proceeds to step S16-4. If an image other than the highlight display is designated as an album image, the process proceeds to step S16-5. For example, when the image 36 in FIG. 17 is selected, the process proceeds to step S16-5.

  In step S16-4, the image processing adaptation history is added. The image processing adaptation history of dictionary data adapted to the priority image selected and designated as the album image is added. For example, if the priority image 34 in FIG. 17 is the adaptation result of the dictionary number 4 dictionary data in FIG. 14, the learning adaptation number is incremented by one. In step S16-5, the image processing adaptation history of the dictionary learning data is subtracted. For example, when a non-priority image of 36 in FIG. 17 is selected, the image processing adaptation history of the dictionary data 4 having the same classification condition as that image is subtracted. For example, 1 is subtracted from the adaptive number.

  In step S16-6, a process of creating new learning data with the selected image is performed. In the example of FIG. 17, since 36 is selected, new dictionary data is created based on the 36 images. In the dictionary data creation process itself, since the classification conditions have already been determined, the selection conditions may be created according to the processing procedure described in the first embodiment based on comparison with the non-selected images.

  In step S16-7, the learning process ends. Open the work area and close the dictionary file with the updated image processing adaptation history. With this configuration, it is possible to realize an image information processing apparatus that determines whether to apply a learning result according to a learning state, and highlights an image having the same tendency as an image that is often selected.

  Each means and each step constituting the embodiment of the present invention described above can be realized by executing a program stored in a RAM, a ROM, or the like of a computer. This program and a computer-readable recording medium recording the program are included in the present invention.

  1: Image information processing device, 2: Pen

Claims (7)

Image selection instruction means for instructing an arbitrary image;
Image classification means for classifying selected images based on classification conditions;
Extraction means for extracting selection criteria for each classification;
Means for storing the classification conditions and selection criteria;
Means for adapting image processing based on stored classification conditions, selection criteria;
An image information processing apparatus comprising:   2. The image information processing apparatus according to claim 1, wherein the means for adapting the image processing automatically arranges images based on classification conditions and selection criteria.   The image information processing apparatus according to claim 1, wherein the means for adapting the image processing displays a list of images based on classification conditions and selection criteria.   The image information processing apparatus according to claim 1, wherein the classification condition is image metadata or image similarity.   The image information processing apparatus according to claim 1, wherein the selection criterion is an image contrast value, an image brightness, an image focus determination value, and an image level. An image selection instruction process for instructing an arbitrary image;
An image classification process for classifying selected images based on classification conditions;
An extraction process for extracting selection criteria for each classification;
Storing the classification conditions and selection criteria;
Adapting image processing based on stored classification conditions, selection criteria;
An image information processing method comprising: An image selection instruction process for instructing an arbitrary image;
An image classification process for classifying selected images based on classification conditions;
An extraction process for extracting selection criteria for each classification;
Storing the classification conditions and selection criteria;
Adapting image processing based on stored classification conditions, selection criteria;
A program that causes a computer to execute.

Images