JP2007122110A - Image classification device, method and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To classify a plurality of images into a plurality of categories with high accuracy. <P>SOLUTION: A classification part 30 calculates a photographing time difference between the adjacent images and an average value of all the photographing time differences when arranging the plurality of images in ascending order or descending order of a photographing date, compares all the photographing time differences and a threshold value obtained from a probability function of the Poisson distribution using the average value, and performs classification processing for classifying an image group comprising the plurality of images into the plurality of categories with betweenness of two images from which the photographing time difference exceeding the threshold value is obtained as a division position of the categories. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an image classification apparatus and method for classifying a plurality of images into a plurality of categories, and a program for causing a computer to execute the image classification method.

  With the widespread use of digital cameras and scanners, anyone can easily acquire digital images. In addition, images are also organized on a personal computer. In this case, images are often organized by classifying images for each event that has occurred to the user. For example, when you go on a trip to Europe, images acquired in France, the United Kingdom, Spain, etc. are classified into categories for each country, and for images classified in the French category, going out in Paris, Nice, etc. Classify images into more detailed categories for each location. For images classified in Paris, the images are classified into finer categories such as the Eiffel Tower, Arc de Triomphe and Notre Dame. Furthermore, regarding Notre Dame, images may be classified into more detailed categories such as in front of the temple, in a remote place, and in the temple.

  If the images are hierarchized and classified into a plurality of categories in this manner, the images divided for each event can be referred to when referring to the images later, so that the images can be easily organized.

  On the other hand, the operation of classifying images into hierarchies is very troublesome for the user because it is necessary to perform the operation while looking at the image and following the user's memory.

For this reason, various methods for automatically classifying images have been proposed. For example, the image is arranged on the time axis based on the information indicating the shooting date and time indicating the shooting date and time given to the image, and the image is compared by comparing the shooting time difference between two images with adjacent shooting times with a threshold value. A method of classifying into k predetermined categories has been proposed (a method of k-means clustering, see Patent Document 1). In addition, the number of images stored based on the shooting date / time information associated with the images is totaled for each predetermined period to obtain the image shooting frequency, and the images were shot within the predetermined period obtained from the shooting frequency. There has also been proposed a method of classifying a plurality of images into a plurality of categories for each event by judging whether the number of images is large or small from a predetermined threshold (see Patent Document 2). Furthermore, a method of classifying images into categories according to the user's schedule by referring to schedule information representing the user's schedule has been proposed (see Patent Document 3).
JP 2000-112997 A JP 2003-141130 A JP 2003-29891 A

  Since the method described in Patent Document 3 needs to manage both the user's schedule and images acquired by photographing, the above-described Patent Document 1 that actually classifies images using only information representing the photographing date and time. , 2 is practical. However, the methods described in Patent Documents 1 and 2 only use the time difference between images with adjacent shooting dates and times, and it is not easy to set an appropriate threshold value. Difficult to do. In addition, since the method described in Patent Document 1 uses the k-means clustering method, it is necessary to determine the value of k, which is the number of categories to be classified, so that the value of k can be set appropriately. If not, there is a problem that the accuracy of classification becomes extremely poor.

  In addition, the method described in Patent Literature 2 has difficulty in defining an appropriate predetermined period, and cannot accurately classify images if the predetermined period is incorrectly determined. Also, the shooting frequency depends not only on the event to be shot but also on the photographer. That is, some photographers shoot every 10 seconds even for the same event, while others shoot every 10 minutes. In such a case, images cannot be classified with high accuracy by the method described in Patent Document 2.

  The present invention has been made in view of the above circumstances, and an object thereof is to allow a plurality of images to be classified into a plurality of categories with high accuracy.

A first image classification device according to the present invention is an image classification device that classifies an image group composed of a plurality of images to which shooting date and time information is assigned into a plurality of categories.
When the image groups are arranged in ascending or descending order of shooting date and time, a shooting time difference between adjacent images and an average value of all the shooting time differences are calculated, and a Poisson distribution using all the shooting time differences and the average value A classification that classifies the image group into the plurality of categories, with a threshold value obtained from the probability function of the image and a difference between the two images having the imaging time difference exceeding the threshold as the category separation position. It is characterized by having a classification means for processing.

  Here, the “probability function of Poisson distribution” is a function for calculating the probability of event A occurring x times within a unit time when there is an event A occurring on average λ times within the unit time. It will be described later.

  In the first image classification device according to the present invention, when the average value is calculated by the classifying unit, if the imaging time difference exceeds a predetermined setting value, the imaging time difference is set as the setting value. It is good also as a means to substitute by a value.

  In the first image classification apparatus according to the present invention, the classification unit may be a unit that repeatedly performs the classification process for the group of images for each classified category every time the classification process is performed.

  Further, in the first image classification device according to the present invention, the classification means may be configured such that the threshold value in the image group or the image group for each classified category is equal to or less than a predetermined minimum value. It is good also as a means to complete | finish the said classification process.

A second image classification device according to the present invention is an image classification device that classifies an image group consisting of a plurality of images to which shooting date and time information is given into a plurality of categories.
When the image groups are arranged in ascending or descending order of shooting date and time, the shooting time difference between adjacent images and the average value of all the shooting time differences are calculated, and the difference between all the shooting time differences and the average value is calculated as described above. Two values obtained by comparing the normalized value based on the variance or standard deviation of all the photographing time differences with the threshold value based on the variance or the standard deviation and obtaining the normalized value exceeding the threshold value The image processing apparatus includes a classification unit that performs a classification process of classifying the image group into the plurality of categories with the image as the category separation position.

  In the second image classification device according to the present invention, when the average value is calculated by the classification unit, if the imaging time difference exceeds a predetermined setting value, the imaging time difference is set as the setting value. It is good also as a means to substitute by a value.

  In the second image classification device according to the present invention, the classification unit may be a unit that repeatedly performs the classification process on the grouped image groups for each classified category every time the classification process is performed.

A third image classification device according to the present invention is an image classification device that classifies an image group consisting of a plurality of images to which shooting date and time information is given into a plurality of categories.
When the image groups are arranged in ascending order or descending order of shooting date / time, a shooting time difference between adjacent images and an average value of all the shooting time differences are calculated, and a weight weighting all the shooting time differences and the average value Classification means for performing a classification process of comparing the group of images into the plurality of categories, with a comparison between the averaged values and the two images obtained with the photographing time difference exceeding the weighted average as the category separation position It is characterized by comprising.

  In the third image classification apparatus according to the present invention, when the average value is calculated by the classification unit, if the imaging time difference exceeds a predetermined setting value, the imaging time difference is set as the setting value. It is good also as a means to substitute by a value.

  In the third image classification apparatus according to the present invention, the classification unit may be a unit that repeatedly performs the classification process on the grouped image group for each classified category.

  Further, in the third image classification device according to the present invention, the classification means may be configured such that the threshold value in the image group or the image group for each classified category is equal to or less than a predetermined minimum value. It is good also as a means to complete | finish the said classification process.

A first image classification method according to the present invention is an image classification method for classifying an image group composed of a plurality of images to which shooting date and time information is given into a plurality of categories.
When arranging the image group in ascending or descending order of shooting date and time, calculate the shooting time difference between adjacent images and the average value of all the shooting time difference,
Comparing all the shooting time differences with a threshold value obtained from a Poisson distribution probability function using the average value,
A classification process for classifying the image group into the plurality of categories is performed with the interval between the two images having the imaging time difference exceeding the threshold as the category separation position.

A second image classification method according to the present invention is an image classification method for classifying an image group consisting of a plurality of images to which shooting date / time information is given into a plurality of categories.
When the image groups are arranged in ascending or descending order of shooting date and time, calculate the shooting time difference between adjacent images and the average value of all the shooting time differences,
Comparing the difference between all the shooting time differences and the average value based on the variance or standard deviation of all the shooting time differences with a threshold value based on the variance or the standard deviation;
A classification process for classifying the image group into the plurality of categories is performed with the interval between the two images obtained with the normalized value exceeding the threshold as the category separation position.

A third image classification method according to the present invention is an image classification method for classifying an image group consisting of a plurality of images to which shooting date / time information is given into a plurality of categories.
When arranging the image group in ascending or descending order of shooting date and time, calculate the shooting time difference between adjacent images and the average value of all the shooting time difference,
Comparing all the shooting time differences with a weighted average value obtained by weighting the average value;
A classification process for classifying the image group into the plurality of categories is performed with the interval between the two images obtained with the photographing time difference exceeding the weighted average value as the category separation position.

  In addition, you may provide as a program for making a computer perform the 1st-3rd image classification method by this invention.

  According to the first image classification apparatus and method of the present invention, when an image group composed of a plurality of images is arranged in ascending or descending order of shooting date and time, the shooting time difference between adjacent images and the average value of all shooting time differences Is calculated. Then, all the shooting time differences are compared with the threshold value obtained from the Poisson distribution probability function using the average value, and the interval between the two images that have obtained an absolute value exceeding the threshold value is set as the category separation position. The image group is classified into a plurality of categories.

  As described above, in the first image classification apparatus and method, the threshold value obtained from the Poisson distribution probability function is used as a threshold value for determining the category separation position. The threshold value is not a constant threshold value but a threshold value adaptively changed according to the actual photographing time difference of images included in the image group to be classified. Therefore, it is possible to classify the image group into a plurality of categories with high accuracy according to the actual photographing time difference between the plurality of images included in the image group.

  According to the second image classification apparatus and method of the present invention, when an image group composed of a plurality of images is arranged in ascending order or descending order of shooting dates and times, all shooting time differences and all shooting time differences between adjacent images are calculated. An average value is calculated. The difference between all shooting time differences and the average value is normalized by the standard deviation of all shooting time differences, and the threshold value based on variance or standard deviation is compared, and the normalized value that exceeds the threshold value The two images obtained are set as category separation positions, and the image group is classified into a plurality of categories.

  As described above, in the second image classification apparatus and method, the difference between all the photographing time differences and the average value is normalized based on the variance or standard deviation of all the photographing time differences, and the variance or standard deviation is used. Since the threshold value is compared, the comparison with the threshold value for classification is performed in a statistically normalized state. Therefore, since the images are classified based on the threshold value adaptively changed according to the normal distribution of the photographing time difference instead of the constant threshold value, the actual photographing time difference of the image group to be classified is Accordingly, the image group can be accurately classified into a plurality of categories.

  According to the third image classification apparatus and method of the present invention, when an image group composed of a plurality of images is arranged in ascending or descending order of shooting dates and times, all shooting time differences and all shooting time differences between adjacent images are calculated. An average value is calculated. Then, all the shooting time differences are compared with a weighted average value obtained by weighting the average value, and two images that have obtained a shooting time difference exceeding the weighted average value are set as a category separation position, and a plurality of image groups are provided. It is classified into the category.

  As described above, in the third image classification apparatus and method, since the weighted average value of all the photographing time differences is used as a threshold value for determining the category separation position, the threshold value for classifying the images. Is not a fixed threshold value, but a threshold value adaptively changed in accordance with the actual photographing time difference of images included in the image group to be classified. Therefore, it is possible to classify the image group into a plurality of categories with high accuracy according to the photographing time difference of the image group to be classified.

  Also, when obtaining the average value, if the shooting time difference exceeds the preset value, the shooting time difference is replaced with the set value, so that the average value and average value of the shooting time difference are used due to the meaningless large shooting time difference. Since it is possible to prevent fluctuations in the threshold value obtained from the Poisson distribution random variable and the weighted average value obtained by weighting the average value, it is possible to classify the images with higher accuracy.

  Further, by repeatedly performing the classification process on the classified image group for each grouped image every time the classification process is performed, the images can be classified in a hierarchical manner.

  Also, the classification process is terminated when the threshold value and the weighted average value obtained from the Poisson distribution probability function in the image group or the image group for each classified category are equal to or less than a predetermined minimum value. Thus, it is possible to prevent two images having a short photographing time difference from being classified into different categories.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a schematic block diagram showing the configuration of an image classification apparatus according to the first embodiment of the present invention. As shown in FIG. 1, the image classification apparatus 1 according to the first embodiment performs various controls such as image data recording control and display control, and also controls each part of the apparatus 1 and operates the CPU 12. A system memory 14 including a program to be executed, an image classification program to be described later, a viewer software for viewing images, a ROM in which various constants are recorded, and a RAM serving as a work area when the CPU 12 executes processing; An input unit 16 including a keyboard and a mouse for performing various instructions to the apparatus 1 and a display unit 18 including a liquid crystal monitor for performing various displays are provided.

  The image classification apparatus 1 also includes a card slot 20 for reading image data from the memory card 2 on which the image data is recorded, and recording image data on the memory card 2, and a technique represented by JPEG. A compression / decompression unit 22 that compresses or decompresses compressed image data; a hard disk 24 that records image data and various programs executed by the CPU 12; and a memory control that controls the system memory 14, the card slot 20, and the hard disk 24 Unit 26 and a display control unit 28 for controlling the display of the display unit 18.

  The image classification device 1 also includes a classification unit 30 that classifies images stored in a processing target folder desired by the user into a plurality of categories among a plurality of folders created on the hard disk 24.

  Hereinafter, an image classification process performed by the classification unit 30 in the first embodiment will be described. In the present embodiment, it is assumed that a plurality of images are stored in the processing target folder. FIG. 2 is a flowchart of the classification process in the first embodiment. The classification unit 30 starts processing when the user instructs image classification from the input unit 16, and first sorts the images stored in the processing target folder in ascending order of shooting date and time (step ST1). It is determined whether there are three or more images stored in the folder (step ST2).

  If step ST2 is negative, that is, if only two images are stored in the processing target folder, a two-image classification process described later is performed (step ST3). If step ST2 is affirmed, a process of calculating an average photographing time difference that is an average value of photographing time differences between adjacent images with photographing dates and times is performed (step ST4). Hereinafter, the average time difference calculation process will be described.

  FIG. 3 is a flowchart of the average time difference calculation process. First, the image to be processed is set to the image with the oldest shooting date (i = 1, step ST21), and the total value of the shooting time differences is set to 0 (step ST22). Then, it is determined whether the photographing time difference between the processing target image and the next image with the next photographing date and time is larger than a preset setting value B0 (step ST23). If step ST23 is affirmed, the new total value is set to a value obtained by adding the set value B0 to the current total value (step ST24). If step ST23 is negative, the new total value is set to a value obtained by adding the photographing time difference to the current total value (step ST25).

  Then, the processing target image is set as the next image (step ST26), and it is determined whether or not the next image exists in the newly set processing target image (step ST27). If step ST27 is affirmed, the process returns to step ST23, and the processes after step ST23 are repeated. If step ST27 is negative, the average photographing time difference is calculated by dividing the total value by the value obtained by subtracting 1 from the total number of images stored in the processing target folder (step ST28), and the average photographing time difference calculating process. Exit.

  Returning to FIG. 2, following step ST4, a threshold value Th1 is calculated from a random variable of Poisson distribution (step ST5). Hereinafter, calculation of the threshold value Th1 will be described.

Here, the “probability function of Poisson distribution” is a function for calculating the probability P (x) of event A occurring x times within a unit time when there is event A occurring λ times within the unit time. Yes, it is represented by the following formula (1).

From the equation (1), when there is a person who takes an image of λ times on average per minute, a function for obtaining the probability P that the image is not taken once in the same event in t minutes is obtained by the following equation (2) )

Here, if the probability P (0) is set in advance and λ is calculated from the photographing time difference between images, t can be obtained by the following equation (3).

  In the present embodiment, the obtained value of t is assumed to be a “threshold value obtained from a Poisson distribution probability function”. Specifically, the value of the probability P (0) in the equations (1) to (3) is set to a value of about 10%, and the number of times of photographing λ per unit time is calculated from the average time difference calculated in step ST4. By doing so, the value of t can be calculated, and the calculated value of t is set as the threshold value Th1.

  Subsequently, it is determined whether or not the calculated threshold value Th1 is larger than a predetermined minimum threshold value Tmin0 (step ST6). If step ST6 is affirmed, the image to be processed is set to the image with the oldest shooting date (i = 1), the category number for classifying the image is set to 1, and the category number of the image to be processed is set to the current number. The category number is set (step ST7). Then, it is determined whether the photographing time difference between the processing target image and the next image with the next photographing date and time is larger than the threshold value Th1 (step ST8). If step ST8 is positive, a new category number is set to a value obtained by adding 1 to the current category number (step ST9), and the category number of the next image is set to the current category number (step ST10). If step ST8 is negative, the process proceeds to step ST10, where the category number of the next image is set to the current category number.

  Subsequently, the processing target image is set as the next image (step ST11), and it is determined whether or not the next image exists in the newly set processing target image (step ST12). When step ST12 is affirmed, the process returns to step ST8, and the processes after step ST8 are repeated. If step ST12 is negative, the process ends.

  On the other hand, if step ST6 is negative, the category number of all images stored in the processing target folder is set to 1 (step ST13), and the process is terminated. In this case, images stored in the processing target folder are not classified at all.

  Next, the two-image classification process will be described. FIG. 4 is a flowchart of the two-image classification process. First, it is determined whether or not the difference in shooting time between two images is greater than a predetermined threshold minimum value Tmin0 (step ST31). If step ST31 is positive, the category number of the first image with the shooting date and time is set to 1 and the category number of the other image is set to 2 (step ST32), and the process is terminated. If step ST31 is negative, the category number of both images is set to 1 (step ST33), and the process ends.

  In addition, after classifying all the images stored in the processing target folder as described above, the classification unit 30 creates a new folder for each category, and stores the images classified in each folder.

  Further, the classification unit 30 may further perform classification processing using the images classified into the folders as processing target folders. The image classification process in each folder is the same as the process shown in FIG. This classification process may be repeated until step ST6 in FIG. 2 is denied, or may be repeated a predetermined number of times. In this way, by repeatedly performing the classification process, it is possible to classify the images stored in the processing target folder as shown in FIG.

  As described above, in the first embodiment, the threshold Th1 obtained from the Poisson distribution probability function is used as a threshold for determining the category separation position. The value is not a constant threshold value, but a threshold value adaptively changed according to a difference in shooting time between a plurality of images stored in the processing target folder. Therefore, it is possible to classify a plurality of images into a plurality of categories with high accuracy according to a difference in shooting time between a plurality of images to be classified.

  Further, in the average photographing time difference calculation process, when the photographing time difference exceeds the preset setting value B0, the photographing time difference is replaced by the setting value B0, so that the meaningless large photographing time difference causes a random variable of the Poisson distribution. Since it is possible to prevent the fluctuation of the obtained threshold value Th1 from becoming large, it is possible to classify the images with higher accuracy.

  In addition, by repeatedly performing the classification process on the classified image group for each grouped image every time the classification process is performed, the images can be classified in a hierarchical manner.

  In addition, when the threshold value Th1 obtained from the Poisson distribution probability function is equal to or smaller than a predetermined minimum value Tmin0, the classification process is ended, so that two images having a short photographing time difference are classified into different categories. Can be prevented.

  Next, a second embodiment of the present invention will be described. FIG. 6 is a schematic block diagram showing the configuration of the image classification apparatus according to the second embodiment of the present invention. In the second embodiment, the same components as those in the first embodiment are denoted by the same reference numerals, and detailed description thereof is omitted here. The image classification device 1 ′ according to the second embodiment includes a classification unit 32 that performs classification processing different from the classification unit 30 in the image classification device 1 according to the first embodiment.

  Next, image classification processing performed by the classification unit 32 in the second embodiment will be described. In the present embodiment, it is assumed that a plurality of images are stored in the processing target folder. FIG. 7 is a flowchart of classification processing in the second embodiment. The classification unit 32 starts processing when the user instructs image classification from the input unit 16, and first sorts the images stored in the processing target folder in ascending order of shooting date and time (step ST41). It is determined whether there are three or more images stored in the folder (step ST42).

If step ST42 is negative, that is, if only two images are stored in the processing target folder, the two-image classification process shown in FIG. 4 is performed (step ST43). If step ST42 is positive, the average photographing time difference calculation process shown in FIG. 3 is performed (step ST44). Subsequently, the standard deviation of the photographing time difference is calculated (step ST45). The standard deviation is calculated by the following formula (4). In Equation (4), σ is the standard deviation, n is the number of images stored in the folder to be processed −1, xm is the average photographing time difference, and xj is the photographing time difference.

  Subsequently, the image to be processed is set to the image with the oldest shooting date (i = 1), the category number for classifying the image is set to 1, and the category number of the image to be processed is set to the current category number. (Step ST46). Then, a value (| xj−xm | / σ) obtained by normalizing the absolute value of the photographing time difference between the processing target image and the next image with the next oldest photographing date by the standard deviation is determined based on the standard deviation. It is determined whether it is larger than Th2 (step ST47). Here, a value such as σ or 2σ may be used as the value of the threshold Th2.

  If step ST47 is affirmed, the new category number is set to a value obtained by adding 1 to the current category number (step ST48), and the category number of the next image is set to the current category number (step ST49). If step ST47 is negative, the process proceeds to step ST49, where the category number of the next image is set to the current category number.

  Subsequently, the processing target image is set as the next image (step ST50), and it is determined whether or not the next image exists in the newly set processing target image (step ST51). If step ST51 is affirmed, the process returns to step ST47, and the processes after step ST47 are repeated. If step ST51 is negative, the process ends.

  As described above, in the second embodiment, the value obtained by normalizing the difference between all the photographing time differences and the average photographing time difference by the standard deviation is compared with the threshold value Th2 determined based on the standard deviation. Therefore, the comparison with the threshold value for performing classification is performed in a statistically normalized state. Therefore, the images are classified not by a fixed threshold value but by a threshold value adaptively changed according to the normal distribution of the photographing time difference. Therefore, it is possible to classify a plurality of images into a plurality of categories with high accuracy according to a difference in shooting time between a plurality of images to be classified.

  Also, in the average shooting time difference calculation process, when the shooting time difference exceeds the preset setting value B0, the shooting time difference is replaced by the setting value B0, so that the average shooting time difference or even the standard is increased due to a meaningless large shooting time difference. Since it is possible to prevent the variation in deviation from becoming large, it is possible to classify images with higher accuracy.

  In addition, by repeatedly performing the classification process on the classified image group for each grouped image every time the classification process is performed, the images can be classified in a hierarchical manner.

In the second embodiment, the value (| xj−xm | / σ) obtained by normalizing the absolute value of the photographing time difference between the processing target image and the next image having the next oldest photographing date and time by the standard deviation is Although it is determined whether or not the threshold value Th2 is larger than the threshold value Th2 determined based on the standard deviation, the absolute value of the photographing time difference between the processing target image and the next image with the next photographing date and time is normalized by the variance σ ² It may be determined whether or not the obtained value (| xj−xm | / σ ² ) is larger than a threshold value determined based on the variance, the deviation value of the photographing time difference is obtained, and the obtained deviation value is used as the standard deviation. Whether or not the threshold value is larger than a predetermined threshold may be determined based on the above.

  Next, a third embodiment of the present invention will be described. FIG. 8 is a schematic block diagram showing the configuration of an image classification apparatus according to the third embodiment of the present invention. Note that in the third embodiment, the same reference numerals are assigned to the same configurations as those in the first embodiment, and detailed description thereof is omitted here. The image classification device 1 ″ according to the third embodiment includes a classification unit 34 that performs a classification process different from the classification unit 30 in the image classification device 1 according to the first embodiment.

  The image classification process performed by the classification unit 34 in the third embodiment will be described below. In the present embodiment, it is assumed that a plurality of images are stored in the processing target folder. FIG. 9 is a flowchart of classification processing in the third embodiment. The classification unit 34 starts processing when the user instructs image classification from the input unit 16, and first sorts the images stored in the processing target folder in ascending order of shooting date and time (step ST61). It is determined whether there are three or more images stored in the folder (step ST62).

  If step ST62 is negative, that is, if only two images are stored in the processing target folder, the two-image classification process shown in FIG. 4 is performed (step ST63). If step ST62 is affirmed, an average photographing time difference calculation process shown in FIG. 3 is performed (step ST64). Subsequently, it is determined whether or not the calculated average photographing time difference (Ts0) is larger than a predetermined minimum threshold value Tmin1 (step ST65). If step ST65 is positive, the image to be processed is set to the image with the oldest shooting date, the category number for classifying the image is set to 1, and the category number of the image to be processed is set to the current category number. (Step ST66). Then, the threshold Th3 (= α ×), in which the difference in shooting time between the processing target image and the next image with the next shooting date and time is an average shooting time difference weighted by a weighting coefficient α (α is a value determined empirically). It is determined whether it is larger than (Ts0) (step ST67). If step ST67 is positive, a new category number is set to a value obtained by adding 1 to the current category number (step ST68), and the category number of the next image is set to the current category number (step ST69). If step ST67 is negative, the process proceeds to step ST69, where the category number of the next image is set to the current category number.

  Subsequently, the processing target image is set as the next image (step ST70), and it is determined whether or not the next image exists in the newly set processing target image (step ST71). If step ST71 is affirmed, the process returns to step ST67, and the processes after step ST67 are repeated. If step ST71 is negative, the process ends.

  On the other hand, if step ST65 is negative, the category number of all images stored in the folder to be processed is set to 1 (step ST72), and the process ends.

  As described above, in the third embodiment, since the weighted average photographing time difference is set as the threshold value Th3 for determining the category separation position, the threshold value for classifying the images is constant. It is not a threshold value but a threshold value that is adaptively changed according to a difference in shooting time between a plurality of images stored in a processing target folder. Therefore, it is possible to classify the image group into a plurality of categories with high accuracy according to the difference in shooting time between the plurality of images to be classified.

  Further, in the average photographing time difference calculation process, when the photographing time difference exceeds the preset setting value B0, the photographing time difference is replaced with the setting value B0, so that the fluctuation of the average photographing time difference due to the meaningless large photographing time difference, Furthermore, since it is possible to prevent fluctuations in the threshold value Th3 from becoming large, it is possible to classify images with higher accuracy.

  In addition, by repeatedly performing the classification process on the classified image group for each grouped image every time the classification process is performed, the images can be classified in a hierarchical manner.

  Further, when the average photographing time difference Ts0 is equal to or smaller than a predetermined minimum value Tmin1, it is possible to prevent the images from being classified into different categories when the photographing time difference is short by ending the classification process. it can.

  In the first to third embodiments, it is determined whether or not three or more images are stored in the processing target folder. If this determination is negative, the two-image classification process is performed. If it is known in advance that there are two images, the two-image classification process may be performed immediately without performing this determination. Further, a classification unit dedicated to the two-image classification process may be provided.

  In the first to third embodiments, in the average photographing time difference calculation process, it is determined whether or not the photographing time difference is larger than a preset setting value B0. If this determination is affirmative, Although the average photographing time difference is calculated by replacing the photographing time difference with the set value B0, the average photographing time difference may be calculated using only the photographing time difference without performing this process.

  In addition, when calculating the average photographing time difference in the first to third embodiments, an image having the latest photographing date and time and an image having the oldest photographing date and time saved in the folder to be processed or the folder classified into the category are used. May be calculated by dividing the difference in shooting time by (number of images-1).

  Although the apparatus according to the first to third embodiments of the present invention has been described above, the computer is caused to function as means corresponding to the above-described classification units 30, 32, and 34, and FIGS. A program for performing the processing shown in FIG. 9 is also one embodiment of the present invention. A computer-readable recording medium in which such a program is recorded is also one embodiment of the present invention. Further, such a program may be incorporated in viewer software for browsing images.

  FIG. 10 is a diagram illustrating an example of a display screen of viewer software in which a program for performing the processing performed by the classification units 30, 32, and 34 is incorporated. As shown in FIG. 10, the screen 50 includes a folder display area 50A for displaying a tree structure of folders, a thumbnail image display area 50B for displaying a list of thumbnail images of images stored in a folder designated by the user, A button display area 50C for displaying buttons for performing various processes is included.

  Three folders 01, 02 and 03 are displayed in the folder display area 50A. It is assumed that a plurality of images are stored in each folder. The button display area 50C includes a classification button 50D for executing a classification process in addition to an enlargement button for enlarging an image and a print button for performing printing.

  Then, after the user designates the folder to be processed in the folder display area 50A, the viewer software executes the classification process by clicking the classification button 50D. For example, if the classification button 50D is clicked after specifying the 01 folder, the images stored in the 01 folder are classified. As a result, the images stored in the 01 folder are hierarchized and classified, and the hierarchized folders are displayed in the folder display area 50A as shown in FIG.

1 is a schematic block diagram showing the configuration of an image classification device according to a first embodiment of the present invention. Flow chart of classification processing in the first embodiment Flow chart of average shooting time difference calculation processing Flow chart of 2-image classification processing The figure which shows the state which hierarchized and classified the image preserved in the processing target folder The schematic block diagram which shows the structure of the image classification device by the 2nd Embodiment of this invention. Flowchart of classification processing in the second embodiment The schematic block diagram which shows the structure of the image classification device by the 2nd Embodiment of this invention. Flowchart of classification processing in the second embodiment Figure showing an example of the display screen of the viewer software (Part 1) An example of the display screen of the viewer software (Part 2)

Explanation of symbols

1,1 ', 1 "Image classification device 2 Memory card 12 CPU
14 system memory 16 input unit 18 display unit 20 card slot 22 compression / decompression unit 24 hard disk 26 memory control unit 28 display control unit 30, 32, 34 classification unit

Claims (17)

In an image classification device for classifying an image group consisting of a plurality of images to which shooting date / time information is given into a plurality of categories,
When the image groups are arranged in ascending or descending order of shooting date and time, a shooting time difference between adjacent images and an average value of all the shooting time differences are calculated, and a Poisson distribution using all the shooting time differences and the average value A classification that classifies the image group into the plurality of categories, with a threshold value obtained from the probability function of the image and a difference between the two images having the imaging time difference exceeding the threshold as the category separation position. An image classification apparatus comprising classification means for performing processing.   The classification means is means for replacing the photographing time difference with the set value when the photographing time difference exceeds a predetermined set value when calculating the average value. The image classification apparatus according to 1.   The image classification apparatus according to claim 1, wherein the classification unit is a unit that repeatedly performs the classification process on the grouped image group every time the classification process is performed.   The classification means is means for terminating the classification process when the threshold value in the image group or the image group for each classified category is equal to or less than a predetermined minimum value. The image classification device according to any one of claims 1 to 3. In an image classification device for classifying an image group consisting of a plurality of images to which shooting date / time information is given into a plurality of categories,
When the image groups are arranged in ascending or descending order of shooting date and time, the shooting time difference between adjacent images and the average value of all the shooting time differences are calculated, and the difference between all the shooting time differences and the average value is calculated as described above. Two values obtained by comparing the normalized value based on the variance or standard deviation of all the photographing time differences with the threshold value based on the variance or the standard deviation and obtaining the normalized value exceeding the threshold value An image classification apparatus comprising classification means for performing classification processing for classifying the image group into the plurality of categories, with an image as the category separation position.   The classification means is means for replacing the photographing time difference with the set value when the photographing time difference exceeds a predetermined set value when calculating the average value. 5. The image classification device according to 5.   7. The image classification apparatus according to claim 5, wherein the classification unit is a unit that repeatedly performs the classification process on the grouped image group every time the classification process is performed. In an image classification device for classifying an image group consisting of a plurality of images to which shooting date / time information is given into a plurality of categories,
When the image groups are arranged in ascending order or descending order of shooting date / time, a shooting time difference between adjacent images and an average value of all the shooting time differences are calculated, and a weight weighting all the shooting time differences and the average value Classification means for performing a classification process of comparing the group of images into the plurality of categories, with a comparison between the averaged values and the two images obtained with the photographing time difference exceeding the weighted average as the category separation position An image classification apparatus comprising:   The classification means is means for replacing the photographing time difference with the set value when the photographing time difference exceeds a predetermined set value when calculating the average value. 8. The image classification device according to 8.   10. The image classification apparatus according to claim 8, wherein the classification unit is a unit that repeatedly performs the classification process on the grouped image group every time the classification process is performed.   The classifying means is means for ending the classification process when the threshold value in the image group or the image group for each classified category is equal to or less than a predetermined minimum value. The image classification device according to claim 8. In an image classification method for classifying an image group consisting of a plurality of images to which shooting date / time information is given into a plurality of categories,
When arranging the image group in ascending or descending order of shooting date and time, calculate the shooting time difference between adjacent images and the average value of all the shooting time difference,
Comparing all the shooting time differences with a threshold value obtained from a Poisson distribution probability function using the average value,
An image classification method comprising: performing classification processing for classifying the image group into the plurality of categories, with two images obtained with the photographing time difference exceeding the threshold as the category separation position. In an image classification method for classifying an image group consisting of a plurality of images to which shooting date / time information is given into a plurality of categories,
When arranging the image group in ascending or descending order of shooting date and time, calculate the shooting time difference between adjacent images and the average value of all the shooting time difference,
Comparing the difference between all the shooting time differences and the average value based on the variance or standard deviation of all the shooting time differences with a threshold value based on the variance or the standard deviation;
An image classification method comprising: performing a classification process for classifying the image group into the plurality of categories, with the interval between the two images obtained with the normalized value exceeding the threshold as the category separation position. In an image classification method for classifying an image group consisting of a plurality of images to which shooting date / time information is given into a plurality of categories,
When arranging the image group in ascending or descending order of shooting date and time, calculate the shooting time difference between adjacent images and the average value of all the shooting time difference,
Comparing all the shooting time differences with a weighted average value obtained by weighting the average value;
An image classification method, wherein classification processing is performed for classifying the image group into the plurality of categories, with two images obtained with the photographing time difference exceeding the weighted average value as the category separation position. In a program for causing a computer to execute an image classification method for classifying an image group composed of a plurality of images to which shooting date / time information is assigned into a plurality of categories,
A procedure for calculating a difference in shooting time between adjacent images and an average value of all the shooting time differences when the image groups are arranged in ascending or descending order of shooting date and time,
Comparing all the imaging time differences with a threshold value obtained from a probability function of a Poisson distribution using the average value;
And a procedure for performing a classification process for classifying the image group into the plurality of categories with the interval between the two images having the imaging time difference exceeding the threshold as the category separation position. In a program for causing a computer to execute an image classification method for classifying an image group composed of a plurality of images to which shooting date / time information is assigned into a plurality of categories,
A procedure for calculating a difference in shooting time between adjacent images and an average value of all the shooting time differences when the image groups are arranged in ascending or descending order of shooting date and time,
Comparing a difference between all the photographing time differences and the average value based on a variance or standard deviation of all the photographing time differences and a threshold based on the variance or the standard deviation;
And performing a classification process for classifying the group of images into the plurality of categories, with the interval between the two images that have obtained the normalized value exceeding the threshold as the category separation position. program. In a program for causing a computer to execute an image classification method for classifying an image group composed of a plurality of images to which shooting date / time information is assigned into a plurality of categories,
A procedure for calculating a difference in shooting time between adjacent images and an average value of all the shooting time differences when the image groups are arranged in ascending or descending order of shooting date and time,
A procedure for comparing all the photographing time differences with a weighted average value obtained by weighting the average value;
A program for performing a classification process for classifying the image group into the plurality of categories, with two images obtained with the difference in photographing time exceeding the weighted average value as division positions of the categories. .

Images