JP2000207405A - Method for classifying plural images, classifying device and recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method and a device which objectively and also efficiently classify and rearrange images in the case of selecting one image being optimum to use among many images when the many images whose contents are the same or almost the same but whose photographing and processing conditions such as color and brightness are different are produced through digital camera photographing and image processing. SOLUTION: When plural images are classified in order of lightness, a density histogram about the specific area of each image is produced about an image group being a classification object, for instance, pixels existing on a low density side are selected by only the number of specific pixels from maximum density pixels Dmax as representative pixels Pi and, the pixel density value is defined as a sort key value Ki. The sort key value is calculated for each pixel being a classification object, and pixels are classified in a lightness difference order by mutually comparing these sort key values. It is possible to perform image classification that is more objective and visually faithful by adopting such a method.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to image processing, and more particularly to automatic classification of a plurality of images having the same or similar images but different shooting conditions.

[0002]

2. Description of the Related Art In recent years, with the advent of high-resolution and inexpensive digital cameras and high-density and low-priced recording media such as memories and disk devices, photographers have been required to use a very large number of digital cameras with slightly different conditions for one subject. It is common to take images. In addition, due to the spread of image processing apparatuses, as a result of performing a conversion process on the color, shade, and gradation of one image, a plurality of images having the same or almost the same image contents but different conditions such as color and brightness are obtained. Groups may be obtained. In such a case, it is necessary for the cameraman or the image processing operator to efficiently select the image having the best condition from the many images in order to increase the productivity of the cameraman or the image processing operator. As an image classification method for this purpose, sorting of a plurality of images photographed with respect to one of these subjects using a certain index (brightness, saturation, hue, etc.) as a classification criterion is followed by selection of an optimal image. In many cases, it has a very high effect.

However, conventionally, when a plurality of images obtained by photographing the same object under different conditions are classified, for example, in order of image brightness, the following method is used. (1) A method in which the operator visually compares the images displayed on the monitor. (2) A method in which the operator visually compares the shape of the histogram regarding the pixel density calculated for each image. (3) The density in the designated area of each image. How to compare numerical values such as average

[0004] Also, when a plurality of images are classified based on a hue value and a saturation value of a specific portion of the image, the same method as described above is used.

[0005]

However, these conventional methods have the following problems when, for example, images are classified in order of brightness. (1) The visual comparison of the monitor display image by the operator depends on the display characteristics of the monitor and the sense of the operator.
Lack of objective criteria. (2) In the case of comparing the shapes of the density histograms for each image by visual observation of the operator, it may not be possible to determine which is brighter, or the operator may subjectively determine. (3) When the comparison is performed based on the density average value of the designated area of each image, a case where a visually brighter image is classified as a dark image and a case where the image is clearly classified may occur.

In the case where these images are classified based on the hue value and the saturation value of a specific portion of the image, it is more difficult to visually compare the saturation and the hue as compared with the case of lightness. Therefore, in these cases, the above problem becomes more remarkable.

[0007] The present invention has been made to solve the problem of such a conventional classification processing of a plurality of images.

[0008]

According to a first aspect of the present invention, there is provided a method for classifying a plurality of images based on pixel density values, comprising the steps of preparing a plurality of target images. Creating a density histogram for each of the prepared images; determining a sort key for each of the images from the density histogram for each of the images; and using the sort key determined for each of the images to form each of the images. Classifying.

According to a second aspect of the present invention, there is provided a method for classifying a plurality of images based on a pixel density value, wherein a step of preparing a plurality of target images, a step of determining a comparison area of each image, Creating a density histogram of each image corresponding to the determined comparison area; determining a sort key of each image from the density histogram of each image; and using the sort key determined for each image. Classifying each of the images.

According to a third aspect of the present invention, there is provided a method for classifying a plurality of images according to the first or second aspect,
Instead of the pixel density value according to claim 1 or 2,
The gist is to classify a plurality of images based on a pixel saturation value or a pixel hue value.

The invention according to claim 4 is an apparatus for classifying a plurality of images based on a pixel density value, means for storing a plurality of target images, and creating a density histogram of each of the stored images. Means, means for determining a sort key for each image from the density histogram of each image, and means for classifying each image using the sort key determined for each image.

According to a fifth aspect of the present invention, there is provided a computer-readable recording medium storing a program for causing a computer to perform a classification process of a plurality of images, wherein the program prepares a plurality of target images in the computer. Creating a density histogram for each of the prepared images; determining a sort key for each of the images from the density histogram for each of the images; and using the sort key determined for each of the images. At least the steps of classifying the images.

[0013]

FIG. 1 is a diagram showing a configuration of an image processing apparatus 1 according to one embodiment of the present invention. The image processing apparatus 1 uses a histogram value of pixel density in a specific region of the image to generate a plurality of images obtained as a result of photographing the same target object under different conditions or performing image processing related to color, gradation, and the like. , Can be classified in the order of density.

As shown in FIG. 1, the image processing apparatus 1 has a configuration utilizing a normal computer system.
The image processing apparatus 1 includes a CPU 1 that executes various arithmetic instructions.
1, a memory 12 for storing programs and the like, a dedicated image memory 13 for storing images to be subjected to image classification processing, a fixed disk 14 serving as a nonvolatile storage area, a display unit 15 such as a display, a mouse, a keyboard, a digitizer, etc. An operation instruction unit 16 for an operator to give an operation instruction, and a reading unit 17 for reading a program and various information from the recording medium 9
To the bus line 8 via an interface as appropriate.
To connect to.

The memory 12 serves as a work area for the CPU 11, and also serves as an area for storing various designation information necessary for the operation of the image classification processing, intermediate data for image processing, and the like. The CPU 11 performs a histogram operation and a sort key operation on the target image under the designated conditions, and finally performs a classification process on all the target images.

The program 21 is stored in the fixed disk 14 from the recording medium 9 via the reading unit 17 and is read into the memory 12 when the program 21 is executed. The recording medium 9 includes a magnetic disk, an optical disk,
This is a so-called removable recording medium such as a magneto-optical disk, and is mainly used for transferring the program 21 from another computer to the image processing apparatus.

The fixed disk 14 is used as a recording medium for reading image data to be subjected to image classification processing and the program from the recording medium 9 into the image processing apparatus 1.

The display section 15 displays the image to be processed by the operator to specify which part of the image to be processed is to be the comparison area, and finally displays the image for the operator to confirm the classified image. indicate.

The operation instructing section 16 is a section for receiving an operator's operation instruction necessary for performing image classification processing, and is constituted by an operation input device such as a mouse, a keyboard, a digitizer, or a stylus pen.

FIG. 2 is a block diagram showing the structure of the image processing apparatus 1 in terms of functions. That is, the CPU 11 operates according to the program 21 to play the role of each component shown in FIG.

The management unit 30 manages the execution of the entire image classification process according to the present invention, and performs mutual information exchange and execution instructions with each of the functional units described below.

The operation designating section 31 is a functional section for selecting a target of image classification processing and setting processing conditions. Specifically, the operator selects a target image group by the operation input section 16 and sets a comparison area in the image. It is used for determination, determination of a histogram target range, and the like.

The histogram calculator 32 is a functional unit that calculates a histogram of pixel values in a comparison area of each image to be classified.

The sort key calculation section 33 is a function section that calculates a sort key value based on specified conditions described later using the histogram values obtained for the respective images.

The image classifying unit 34 performs a process of classifying the target image groups in a designated order (for example, in order of increasing density) based on the sort key value calculated for each image. It is. Normally, this processing is realized by changing the pointer indicating the storage position of the image data, instead of replacing the image data itself of the target image group in the storage device.

The display section 15 is used for displaying the image 40 to be processed, for the operator to specify a comparison area in the image to be processed, and for confirming the image after the classification processing.

The configuration of the image processing apparatus for performing the image classification processing has been described above. Next, the flow of the entire image classification processing using this image processing apparatus will be described. FIG. 3 is a diagram showing the outline of the operation of the image processing apparatus 1 and the flow of the work of the operator.

First, in S-10, the management unit 30 determines a group of images to be classified. Here, the operator may directly specify the folder in which the group of classification target images exists, or the management unit 30 searches for the target image based on the information specified by the operator (for example, the file name or the image generation date and time). May be determined. Further, the management unit 30 may automatically determine a plurality of image groups stored in the fixed disk 14 in advance based on a predetermined order and condition.

Next, in S-20, a comparison area in each image of the image group determined in S-10 is determined.
As an example of the specific method, the operator designates the internal area of the target image 40 displayed on the display unit 15 as a closed area such as a rectangle by using a mouse or the like of the operation input unit 16. Alternatively, the management unit 30 may automatically determine the parameter indicating the comparison area with reference to a file or the like in which the parameter is written in advance. This step is skipped when comparing and contrasting all image areas from the beginning.

In step S-30, a target density range R to be used for image classification is determined from a histogram diagram (an example of which is shown in FIG. 5) obtained for each image in step S-60. Parameter values s, d, T required for
Is determined. That is, s and d are the cut rate s indicating the percentage of the total area of the out-of-range area S shown in FIG. 5 and the distribution rate d to the high density area and the low density area. By using these and a determination key T to be described later, a minimum density value D that determines a target density range R for each image in S-60 to be described later.
min and the maximum density value Dmax are first determined. The cut rate s and the distribution rate d are in a low concentration range of Dmin or less,
And a range excluding a high-density range equal to or higher than Dmax is set outside the target density range. The reason that the target density range R used for image classification is limited to the above range is that the low density range and the high density range generally contain a large amount of image noise components, and the image density is classified by including these components. When the processing is performed, when the two images after the classification processing are visually compared, an image recognized as a bright image may result in inconvenience such as being classified as a dark image as a result of the classification processing according to the present invention. That's why. The distribution ratio d is, for example, 60
%, And 40% on the low density side, for a given cut rate s (for example, when s is 20% of the total number of pixels), the high density side of the histogram corresponding to each image is 0.1%. 6s (for example, 12% of the total number of pixels), and 0.
4 s (for example, 8% of the total number of pixels) is cut. Because the noise component of the image is extremely small,
If the above-described inconvenience does not occur even if the image classification processing is performed in the entire image density range without specifying the out-of-range area S, this step can be omitted.

The determination of the target density range R in S-30 is performed by specifying the cut rate s and the distribution rate d, for example, by the operator specifying these density values.

At S-30, a determination key T is determined in addition to the determination of the cut rate s and the distribution rate d. The determination key T is a decimal value of 1 or less used for obtaining a deviation pixel number Tnum, which will be described later, and is a numerical value designating what percentage of a pixel in the target density range R is to be selected as a representative pixel described later. is there. That is, assuming that the number of all pixels included in the target density range is M, Tnum is obtained by the following equation. Tnum = T × M (1) Image classification processing (classification processing based on density value) according to the present invention
Is, as described below, a representative pixel Pi (i =
1 to n; n is the number of images in the target image group), and is compared by using the density value Ki as a sort key. This Tnu
m is the number of the representative pixel Pi of each image counted from the specific density value (Dmax is used in the present embodiment) included in the target density range R (in this embodiment, from the high density pixel). (Counting in the direction of low density).

The process of determining the representative pixel Pi of each image in S-70 will be described first to facilitate understanding.
FIG. 5 shows the density value K of the representative pixel in the present embodiment.
FIG. Cut rate s with S-30
When the distribution ratio d and the determination key T are determined, the histogram diagram corresponding to FIG. 5 and the density values Dmax and Dm at both ends of the target density range are provided for each image in S-60 described later.
in is required. Then, in S-70, first, the total number of pixels M within the target density range R is obtained, and the number of deviation pixels Tnu is determined by the determination key T determined in S-30 based on the equation (1).
Find m. Next, based on the histogram diagram, in the present embodiment, a pixel counted by the pixel number deviation Tnum in the low density direction from the specific density value Dmax is searched, and this is searched for the image (the i-th image in the group of images to be classified). ), And its density value Ki is obtained as a sort key value. The above is the processing in S-70.

Returning to the description of the flowchart of FIG.
In order to easily and sequentially retrieve each image of the target image group determined in -10, an image list including a pointer indicating a storage location of each image data and a connection structure thereof is created. FIG. 6 shows the structural relationship.

Next, in steps S-50 to S-80, a histogram diagram of each image is obtained, and using this, a representative pixel of each image used for image classification is obtained.

In S-50, the first image in the target image group is selected using the list created in S-40. Next, in S-60, the comparison area (S-2
The density values Dmax and Dmin at both ends of the target density range for the pixels in the (area determined by 0) are determined, and the histogram calculation is performed in this range.

Here, the details of the histogram calculation in S-60 will be described with reference to the detailed flowchart of FIG. First, as an initial setting for calculation preparation, S-
At 61, a histogram counter is provided, and its internal value is initialized to zero. Here, the histogram counter is
The number of all density values of the normal image (when the density value of one pixel and one color is represented by 8 bits, normally 256 from 0 to 255)
Means a counter corresponding to. In the present embodiment, the histogram calculation is based on Dmin and Dmax determined in S-60.
Is performed only on the pixels within the range of と, the histogram counter is (Dmax) from Dmin to Dmax.
It is sufficient to prepare -Dmin + 1) counters.

Next, at S-62, the (S-2)
Read the first pixel in the comparison area (determined by 0).
Thereafter, in the loop processing from S-63 to S-66, the images in the comparison area are sequentially read and processed.

In S-63, the pixel value (density value in this embodiment) of the read pixel is obtained, and the histogram counter corresponding to the pixel value is determined. For example, if the density value is 1
If it is 12, the 112th histogram counter is selected. In the next S-64, the histogram counter value corresponding to the density value is advanced by one. Then, in S-65, it is checked whether or not all the pixels in the comparison area have been read out. If No, the pixel value of the next pixel is read out in S-66. By repeating the above processing, a histogram of the pixels in the comparison area is finally obtained in the histogram counter.

Returning to the description of FIG. When the histogram calculation of the target image is performed in S-60 as described above, the next step is S-70.
As described above, the representative pixel Pi of each image is determined, and its pixel value (the density value in the present embodiment) is read out and set as the sort key value Ki. Next, in S-80, it is checked whether or not all the images to be classified have been processed. If No, the process returns to S-50, and the processing of the next image is continued.

On the other hand, if the processing of all the images has been completed, S-9
At 0, the sort key values Ki obtained for the respective images are compared, and are arranged, for example, in ascending order. Then, the sorted list corresponding to this order is performed by changing the list pointers in FIG. If necessary, these classified images are displayed on the display unit 1 in order from the brighter image, for example.
5, the operator is informed of the classification result.

<Other Embodiments> In the above-described embodiments, the criterion for image classification has been described as the pixel density value. However, this may be obtained from the hue value H or the saturation value S of the pixel. Assuming that the RGB values of the target pixel p are R, G, and B, the hue value Hue (p) and the saturation value Sat (p)
When obtaining from the GB value, the following equation is used.

In order to classify the target image group based on the hue value, the pixel values of the target image in FIGS. 3 and 4 are obtained by equation (5) to form a histogram, the representative pixel is determined, and the hue value is determined. May be used as the sort key value to perform the image classification processing. In this case, since the hue value has a value of 0 to 360 degrees, a corresponding histogram counter (for example, quantization to an integer value of 360 from 0 to 359 degrees, 36
(Corresponding to zero counters).

Similarly, when the target image group is classified on the basis of the saturation value, the pixel values of the target image in FIGS. 3 and 4 are obtained by equation (8) to form a histogram, and the representative pixel is determined. Image classification processing may be performed by using the saturation value as a sort key value. In this case, since the hue value has a value from 0 to 1, for example, the obtained saturation value is 0.0039.
It is possible to quantize every (1/256), convert it to a numerical value from 0 to 255, and prepare 256 histogram counters corresponding to this.

[0045]

As described above, according to the first, fourth and fifth aspects of the present invention, a representative pixel is selected based on a histogram of the density and the like for each of the images to be classified, and the density value and the like are sorted. Since image classification is performed as a value, it is possible to perform image classification based on an objective criterion corresponding to the pixel value of each image, which has the effect of enabling accurate and objective image classification that is independent of the operator's subjectivity. Also, using the classification result, the operator can easily select the image or the like having the best brightness.

According to the second aspect of the present invention, since a partial area within each pixel of the classification target image can be designated as a comparison area, for example, the brightness, hue, saturation, etc., of only the person portion of the target image can be specified. Image classification as a reference is possible,
This has the effect that image classification with finer grain is possible.

According to the third aspect of the present invention, the image can be classified based on the hue value or the saturation value of each pixel of the classification target image. This has the effect that an image with a vividness can be easily selected.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating a configuration of an image processing apparatus according to an embodiment.

FIG. 2 is a block diagram illustrating a configuration of the image processing apparatus according to the embodiment from a functional aspect.

FIG. 3 is a diagram showing a flow of an image classification process corresponding to the embodiment.

FIG. 4 is a flowchart showing a procedure for calculating a histogram.

FIG. 5 is a histogram diagram of an image.

FIG. 6 is a diagram exemplifying a structure of a list of images.

[Explanation of symbols]

 Reference Signs List 1 image processing device 9 removable recording medium such as FD 11 CPU 12 memory 13 image memory 14 fixed disk 15 display unit 16 operation instruction unit 17 reading unit 21 program 30 management unit 31 operation designation unit 32 histogram operation unit 33 sort key operation unit 34 image Classification unit 40 Processing target image

Claims (5)

[Claims] 1. A method for classifying a plurality of images based on pixel density values, comprising: preparing a plurality of target images; creating a density histogram of each of the prepared images; An image classification method for a plurality of images, comprising: determining a sort key for each image from the density histogram; and classifying each image using the sort key determined for each image. 2. A method for classifying a plurality of images based on pixel density values, comprising: preparing a plurality of target images; determining a comparison region of each image; Creating a density histogram for each image, determining a sort key for each image from the density histogram for each image, and classifying each image using the sort key determined for each image A method of classifying a plurality of images. 3. A method for classifying a plurality of images according to claim 1 or 2, wherein a pixel saturation value or a pixel hue value is used instead of the pixel density value according to claim 1 or 2. An image classification method characterized by classifying a plurality of images based on the image. 4. An apparatus for classifying a plurality of images based on a pixel density value, comprising: means for storing a plurality of target images; means for creating a density histogram of each of the images; An image classification device, comprising: means for determining a sort key for each image from the density histogram; and means for classifying each image using the sort key determined for each image. 5. A computer-readable recording medium storing a program for causing a computer to perform a classification process on a plurality of images, wherein the program prepares a plurality of target images in the computer, Creating a density histogram of the image, determining a sort key for each image from the density histogram of each image, and classifying each image using the sort key determined for each image, And a computer-readable recording medium for executing at least the following.