JPH0944519A - Method and device for image retrieval - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method and device for image retrieval which can reduced in the process quantity during image retrieval and perform image retrieval from a large quantity of frames at high speed. SOLUTION: A buffer 11 is stored with image data generated by processing respective blocks by DCT transformation and compressing them. A variable length decoding part 2, an inverse quantization part 3, and a DC component separation part 7 separate DC components of the respective blocks from the image data in the buffer 1. A 1st matching part 8 extracts blocks which meet retrieval conditions inputted through a retrieval condition input part 6 on the basis of the DC components. The blocks which are thus extracted are restored to image data in bit map format by an inverse DCT part 4 and a 2nd matching part 5 decides whether or not they meet the retrieval conditions. When it is decided that the restored blocks meet the retrieval conditions, the image is outputted as a retrieval result.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image search method and apparatus for searching a desired image from compressed image data.

[0002]

2. Description of the Related Art Conventionally, as an image compression device, for example, one using a JPEG algorithm for still image compression and an MPEG algorithm for moving image compression is known. These algorithms use DCT (discrete cosine transform)
Is used to map the image to the spatial frequency domain, the amount of data is concentrated in the low frequency components, and the high frequency components are cut to reduce the amount of data.

A block diagram of an image compression apparatus using the MPEG algorithm is shown in FIG.

The image data input by the image input unit 81 is input as it is to the DCT unit 83 in the mode where no prediction is performed. In the prediction mode, the input image data and the predictor 90
The difference from the prediction data is calculated, and the result is calculated by the DCT unit 83.
Enter The predictor 90 includes an inverse quantizer 87 and an inverse D.
The prediction data is output based on the data decoded by the CT unit 88.

In the DCT section 82, the luminance Y and the color difference C
For each component of b and Cr, for each block of a predetermined size (for example, 8
Discrete cosine transform is performed for each block of x8 pixels) and a spatial frequency component as shown in FIG. 10 is obtained for each block.

The quantizing section 84 performs quantization by dividing the value of the spatial frequency component by the value of the quantization table and the quantization coefficient. As shown in FIG. 11, the quantization table defines values corresponding to the spatial frequencies in the block, and generally has a large value as it corresponds to high frequency components. This cuts off high frequency components. Further, one quantized coefficient for controlling the compression rate is defined for each block.

In the variable length coding unit 85, the quantized value is coded by entropy coding while performing zigzag scanning from the low frequency component to the high frequency component, and the coded data is output to the buffer 86. It

The quantization coefficient determination unit 91 reduces the data amount by increasing the value of the quantization coefficient when the amount of data output to the buffer 86 is large, and the value of the quantization coefficient when the amount of data is small. Increase the amount of data by making it smaller. As a result, the amount of data is made as constant as possible regardless of the original image data.

FIG. 9 shows a general configuration of an apparatus for searching a frame matching a desired condition from compressed data of a moving image obtained by the above apparatus.

Usually, when performing a moving image search at high speed,
From the compressed data loaded in the buffer 1, only frames compressed in a mode in which interframe prediction is not performed are selected and then decoded. The mode used for compression is determined by a predetermined flag in the data of each frame.

The data of the frame selected by the buffer 1 is decoded by the variable length decoding unit 2 in the reverse procedure of the variable length coding unit 85. The inverse quantization unit 3 performs inverse quantization by the procedure reverse to that of the quantization unit 84. Inverse DCT unit 4
Then, the inverse discrete cosine transform is calculated by the procedure reverse to that of the DCT unit 83 to restore the image data. The matching unit 5 determines whether this image data matches the search condition input by the search condition input unit 6, and if it matches, the image data is output as a search result.

[0012]

As described above, in the conventional retrieval of compressed images, one frame is restored to the original image format and then the retrieval conditions are matched. Therefore, it is necessary to perform time-consuming inverse DCT calculation and pattern matching in pixel units over the entire frame, and when searching from a large number of frames, the search time becomes enormous. there were.

The present invention has been made in view of the above problems, and provides an image search method and apparatus which can reduce the processing amount at the time of image search and can perform image search from a large number of frames at high speed. The purpose is to provide.

Further, the present invention uses the inverse D in image retrieval.
The purpose is to speed up image retrieval by reducing the number of times CT calculation or pattern matching in pixel units is performed.

[0015]

An image retrieval apparatus of the present invention for solving the above-mentioned problems has the following configuration.
That is, an image retrieval device that divides an image into a plurality of blocks and retrieves a desired image from image data expressing each block with a spatial frequency component, and obtains a predetermined spatial frequency component of each block from the image data. Based on the acquisition means and the predetermined frequency component acquired by the acquisition means,
Extraction means for extracting a block satisfying a preset search condition, restoration means for restoring the block to image data by bitmap based on the spatial frequency component of the block extracted by the extraction means, and the restoration means Determining unit that determines whether or not the restored block satisfies the search condition, and outputs the image as a search result when the determining unit determines that the restored block satisfies the search condition And output means.

Preferably, the predetermined spatial frequency component acquired by the acquisition means is a direct current component. This is because the DC component indicates an average color and brightness in each block, and is suitable for rough search.

Further, preferably, the search condition is a color designated by an operator.

Further, preferably, in order to set the search condition, a display means for displaying a color sample and a designating means for designating a desired color from the color sample displayed on the display means are further provided. This is because it is possible to specify the color to be searched from the displayed color sample, and thus it is possible to specify the color more accurately.

Further, preferably, when the block is extracted by the extracting means, the search result is output by any one of the extracted image output means for outputting the image as a search result, the extracted image output means and the output means. And a switching means for switching between the two. This is because the processing can be switched according to the required search accuracy.

Further, an image retrieval apparatus having another configuration of the present invention for achieving the above object divides an image into a plurality of blocks, and retrieves a desired image from image data expressing each block with a spatial frequency component. An image retrieving apparatus for retrieving a predetermined spatial frequency component of each block from the image data, and a retrieval image and a pattern preset based on the predetermined frequency component obtained by the obtaining unit. Extraction means for performing matching and extracting an area that matches the search image; and restoration means for restoring the area to image data by bitmap based on the spatial frequency component of the area extracted by the extraction means, Determination means for determining whether the area restored by the restoration means matches the search image by pattern matching, and the restoration means for the restoration means If the frequency is determined to match with the search image, and output means for outputting the image as a search result.

Further, preferably, the extracting means includes a first dividing means for dividing the search image into a plurality of color regions formed by colors in a predetermined similar range, and an image formed by the predetermined spatial frequency component. Dividing into a plurality of color regions formed by colors in a predetermined similar range, and the first and second
Comparing means for checking the matching by comparing the colors of the respective areas obtained by the dividing means. Since the extraction is performed by comparing the colors for each color area, the matching process can be performed at high speed.

Further, preferably, the comparing means sets a region having the largest area among the plurality of color regions in the search image as a representative region, and selects from the image composed of the predetermined spatial frequency components as the representative region. When a region having a color in the same similar range is detected, it is compared with another color region of the search image. Color area with the largest area (representative area)
Since the search can be performed to narrow down the initial stage, the extraction process can be performed more efficiently.

Further, preferably, the comparison means further compares the ratio of the size of each color area in the search image with the ratio of the size of each corresponding color area of the image composed of the predetermined spatial frequency component. To do. This is because more accurate pattern matching can be performed.

Further, preferably, when the area is extracted by the extraction means, the search result is output by any one of the extracted image output means for outputting the image as a search result, the extracted image output means and the output means. And a switching means for switching between the two. The processing can be switched according to the required search accuracy, and the search result can be obtained at higher speed when the accuracy is not required.

According to the above arrangement, the image is divided into a plurality of blocks, and in the image retrieval apparatus for retrieving a desired image from the image data expressing each block with the spatial frequency component, a predetermined image of each block is retrieved from the image data. Spatial frequency components are acquired. Then, a block satisfying a preset search condition is extracted based on the acquired predetermined frequency component. The block thus extracted is restored to the image data by the bit map based on the spatial frequency component, and it is further determined whether or not the above search condition is satisfied. When it is determined that the restored block satisfies the search condition, the image is output as the search result.

For example, when a direct current component is used as the predetermined spatial frequency component and a color is designated as the search condition,
The extraction means can make a rough judgment based on the average color of each block. Then, the block matching the search condition can be restored to the original image, and the determination unit can further make a detailed determination based on the color of each pixel.
Since it is possible to narrow down the blocks to be subjected to the detailed determination by the rough determination by the extraction means, the processing efficiency is improved and the search speed is improved.

Further, according to the above-mentioned other structure, the image is divided into a plurality of blocks, and in the image retrieving apparatus for retrieving a desired image from the image data expressing each block by the spatial frequency component, each block is retrieved from the image data. A predetermined spatial frequency component of is acquired. Then, pattern matching with a preset search image is performed based on the acquired predetermined frequency component, and a region that matches the search image is extracted. Then, the extracted area is restored to image data by a bitmap based on the spatial frequency component. Pattern matching is performed on the restored area to determine whether it matches the search image. Then, when it is determined that the restored area matches the search image, the image is output as the search result.

[0028]

BEST MODE FOR CARRYING OUT THE INVENTION Preferred embodiments of the present invention (hereinafter referred to as embodiments) will be described below with reference to the accompanying drawings.

(Embodiment 1) FIG. 1 is a block diagram showing an outline of a hardware configuration of an image search apparatus according to this embodiment. In the figure, 51 is a hard disk,
Image data compressed by DCT coding is stored. A tablet 52 is used as a pointing device for the operator to give a command to the apparatus. Reference numeral 53 denotes an LCD, which displays various information from the image search device to the operator. In this embodiment, LC
A color liquid crystal display is used as D53. 5
An operation unit 4 performs inverse DCT, inverse quantization, and other various operations. 55 is a random access memory (RA
M), and provides a work area for storing data that is being compressed. Reference numeral 56 denotes a control unit, which is the ROM 56a
Various processes are realized by the CPU 56a executing the control program stored in.

FIG. 2 is a block diagram showing the functional arrangement of the image retrieval apparatus according to this embodiment.

From the compressed data loaded from the HDD 51 into the buffer 1, a frame compressed in a mode in which interframe prediction is not performed is selected and transferred to the variable length decoding unit 2. The mode in which the compressed data is compressed can be realized by a well-known technique such as determining by a predetermined flag in the data of each frame. The variable length decoding unit 2 decodes the compressed data transferred from the buffer 1 by a procedure reverse to that of the variable length coding unit 85 shown in FIG.
The inverse quantization unit 3 performs inverse quantization by the procedure reverse to that of the quantization unit 84 shown in FIG. The inversely quantized data represents the spatial frequency component of each block.

As shown in FIG. 10, the component having the lowest spatial frequency is the direct current (DC) component, which corresponds to the average value of the luminance component or the color difference component in the block. Therefore, it can be seen that if the DC component of the DCT result for the luminance component Y is high, it is a bright block, and if it is low, it is a dark block. Further, it can be seen that if the DC component of the DCT result for the color difference component Cb is high and the DC component of the DCT result for the color difference component Cr is low, the block is bluish. Similarly, if the DC component for Cb is low and the DC component for Cr is high, the block is reddish, and if the DC components for Cb and Cr are both high, the block is purpleish, Cb and Cr.
It can be seen that the block is greenish if both the DC components with respect to are low. Therefore, by the DC component separation unit 7,
An outline of the image can be obtained by extracting only the DC components of each block.

The first matching unit 8 determines whether the rough image data composed of the DC component matches the search condition input by the search condition input unit 6, and if they match, the area determination unit. The block of the part matched by 9 is extracted. In the inverse DCT unit 4, the area determining unit 9
The DCT unit 8 shown in FIG. 8 for the blocks extracted in
Inverse Discrete Cosine Transform, which is the reverse procedure of 3, is performed to restore as image data. Then, the second matching unit 5 determines whether or not the image data matches the search condition input by the search condition input unit 6, and if they match, the image data is output as a search result.

The operation of the image search apparatus of the present embodiment having the above configuration will be further described with reference to the flowchart of FIG. FIG. 3 is a flowchart showing an operation procedure of the image search device according to the first embodiment. The control program for realizing the control shown in FIG. 3 is stored in the ROM 56b and executed by the CPU 56a.

In step S101, the operator inputs search conditions from the search condition input unit 6. In this example, a color is specified as the search condition, and it is determined whether or not a part of the input image includes the specified color. FIG. 4 is a diagram showing a display for designating a color as a search condition. Figure 4
As shown in, pointing to one color from the color sample displayed on the screen and selecting with the device.

Next, in step S102, an image composed of DC components is generated from the compressed data loaded in the buffer 1. That is, in step S102, one frame in a mode in which inter-frame prediction is not performed is taken out, variable length decoding unit 2 performs variable length decoding on the frame, and inverse quantization unit 3 further performs inverse quantization. . Then, the DC component separating unit 7 extracts the DC component of each block to form an image.

In step S103, the first matching section 8 determines whether or not the DC component image generated in step S102 matches the search condition input in step S101. Specifically, the color of each pixel of the DC component image is compared with the color selected from the color sample.
The luminance component Y, the color difference component Cb, of a pixel in the DC component image,
When Cr is Yi, Cbi, and Cri, and the luminance component Y and the color difference components Cb and Cr of the color selected from the color sample of FIG. 4 are Ys, Cbs, and Crs, respectively,

[0038]

[Equation 1]

If the value of is within a predetermined value, it is determined that a match has occurred.

In step S104, step S
In the process of 103, it is determined whether or not there is a matched block, and if there is, the process proceeds to step S105. If there is no matched block in step S104, the process proceeds to step S109 to search for the next frame.

In step S105, the matched block is extracted by the area determining section 9, and the inverse DCT section 4 performs an inverse discrete cosine transform operation on the extracted block to restore it as image data. Then, in step S106, the luminance component Yi and the color difference components Cbi, Cri are obtained for each pixel of the block restored in step S105, and the value represented by the equation (1) is calculated as in step S103. . Then, the ratio of the number of pixels whose calculated value is within a predetermined value to the total number of pixels is calculated.

Next, in step S107, it is determined whether or not the frame matches the search condition based on the ratio obtained in step S106. Here, if the ratio obtained in step S106 is greater than or equal to the predetermined ratio, it is determined that the frame matches the search condition. As a result of this determination, if the search condition and the frame match, to step S108; otherwise, step S108.
Go to 109 respectively. In step S108,
The image of that frame is output as the search result. In step S109, it is determined whether or not the matching process has been completed for all the frames, and if there is an unprocessed frame, the process returns to step S102 and the above process is repeated. Also,
If the processing has been completed for all the frames, this processing is ended as it is.

As described above, steps S102 to S10 are performed.
By repeating the operations up to 8 for all the frames, it is possible for the operator to search for the frame that matches the search condition only by inputting the search condition once.

As described above, according to the present embodiment, when orange is designated by the color sample when searching for a person wearing orange clothes, for example, the average color (D
Matching is performed by the image of the C component), and if this is separated from orange in the color space, the block is removed by the first matching unit 8 and the region determination unit 9. The remaining blocks may include, for example, blocks in which red and yellow striped patterns are averaged to become orange. Therefore,
By performing the inverse DCT operation only on the remaining blocks, checking the color of each pixel and performing matching by the second matching unit 5, the red and yellow striped pattern part is removed, and an image composed of orange pixels You can leave. That is, the orange image can be searched without performing the inverse DCT calculation on the entire frame, and the red and yellow images are not searched by mistake.

As described above, in the present embodiment, the target of matching can be narrowed down at the stage of the DC component image, so that the processing amount of inverse DCT conversion and the processing amount of matching can be reduced, and the image search processing can be performed. Processing speed is improved.

In the above embodiment, the matching is determined for each frame for the mode in which the inter-frame prediction is not performed, but the determination may be performed for each predetermined number of frames. Further, the determination may be made every time the scene is switched. As a method of automatically detecting that the scene has been switched, a known method is a method of obtaining a difference from the immediately preceding frame. Since it is highly likely that consecutive frames are similar in a moving image, matching is performed every predetermined number of frames or each scene change to improve practical search accuracy and improve processing speed. be able to.

Although the moving image compression apparatus has been described in the present embodiment, it is possible to retrieve an image of a still image by the same procedure, and the same effect as that applied to a moving image can be obtained.

In this embodiment, DCT, quantization,
Although an example of performing compression by a procedure called entropy coding has been shown, any method may be used as long as it is a compression method that can extract the DC component of the spatial frequency of each block during the expansion of the compressed data. Absent.

(Second Embodiment) Next, a second embodiment will be described. In the second embodiment, the operator specifies a search image, and the search condition is how much the DC component image matches the search image. The hardware configuration and the functional configuration of the image search device of the second embodiment are the same as those of the first embodiment (FIGS. 1 and 2).

The operation of the image search apparatus according to the second embodiment will be described below. The flow of processing is almost the same as that of the flowchart of FIG. 3, but steps S101, S103 and S106 are different from those of the first embodiment. The points different from the first embodiment will be described below.

The method of designating the search image in step S101 is obtained, for example, by selecting one image from the moving images to be searched and cutting out a region with a pointing device, or by other image input means such as an image scanner. A method of using an image, a method of drawing with a pointing device, and the like are possible.

Details of the matching process between the DC component image and the search conditions in steps S103 and S106 will be described with reference to the flowchart of FIG. FIG. 7 is a flowchart showing the procedure of matching processing in the second embodiment.

In step S201, the search image is divided into rough color areas. In the example of FIG. 5, black (area 2
1), flesh color (area 22), yellow (area 23), and blue (area 24). Next, step S20
In 2, the color of the area having the largest area of the color areas is set as the representative color of the search image. In the example of FIG. 5, yellow (area 23) is the representative color.

Here, the color area division is automatically performed. In this division processing, basically, a predetermined number of main colors (main colors) are determined, all the colors of the input image are assigned to any of the main colors described above, and adjacent pixels of the same main color are allocated. Create an area by integrating For example,
If each of the Y, Cb, and Cr components can take a value of 256 gradations, each value is divided by 64 to obtain a value of four stages. 4 × 4 × 4 = 64 with these combinations
It is divided into the main colors of the street. However, since there are combinations that are not possible in reality, the number is actually less than 645. In addition, YCb of major colors such as white, black, red, blue, and yellow
The Cr value may be determined in advance, and each YCbCr value of the input image may be compared and assigned to the closest major color.

When the image is divided into the color areas for each main color as described above, in reality, many small areas are generated in addition to the large area. Therefore, by ignoring small areas that are less than the prescribed size or by combining areas of similar colors, it is possible to divide into areas that only consider large areas. To be done.

In step S203, step S102
The DC component image generated in (FIG. 3) is divided into rough color regions. Then, in step S204,
From the color area of the DC component image, an area that matches the representative color of the search image set in step 202 is extracted. Here, the luminance component Y and the color difference component C of the color region of the DC component image
b and Cr are respectively Yi, Cbi and Cri, and the luminance component Y and the color difference components Cb and Cr of the color region of the search image are Ys and Cri, respectively.
If Cbs and Crs are used, and if the value obtained by the above equation (1) is within a predetermined value, both areas are considered to match.

When a matching area can be extracted from the DC component image, the process proceeds from step S205 to step S206, and the area of the representative color area in the extracted DC component image and the area of the representative color in the search image are extracted. Ask for money. Then, in step S207, matching between the color of each area in the search image and the color of the area around the representative color in the DC component image is checked. At this time, the area around the representative color in the input image is searched based on the area ratio of the areas of the representative color obtained in step S206.

For example, when the search image of FIG. 5 is matched with the DC component image, it is assumed that the area 27 of FIG. 6 matches the representative color (yellow). In this case, it is checked whether the areas around the area 27 match the colors of the areas 21, 22, and 24 in the search image. For example, the representative color area 27 of the DC component image and the representative color area 23 of the search image
Assuming that the area ratio to the area is 1.5: 1, it is first checked whether or not there is a black area having an area about 1.5 times as large as the area 21 at the position 25 in FIG. Next, it is checked whether or not there is a flesh-colored area having an area approximately 1.5 times as large as the area 22 at the position 26. Further, it is examined whether or not there is a blue area having an area of about 1.5 times that of the area 24 at the position 28.

Here, a region having an area of about 1.5 times means a certain range such as 1.3 times to 1.7 times. Increasing this range increases the possibility of searching for unnecessary images, and narrowing the range increases the possibility of missing necessary images. Particularly, in the matching by the DC component image, the error is large because it is configured in block units instead of pixel units, and it is necessary to set the above range to a certain extent.

In step S208, the ratio of the area matching the DC component image to the area of the search image is set to 0.
Output with a numerical value of ~ 1. For example, in the example of FIG. 5, the number of areas of the search image is 4 including the representative color, and 3 of them are included.
If each region matches the DC component image, 0.7
A numerical value of 5 is output as the matching degree. In addition, it goes without saying that the size of the region of each color may be taken into consideration when calculating the matching degree. For example, when the flesh-colored area 26 does not have a size 1.5 times as large as the area 22, the size deviation is reflected in the matching degree. When the color that matches the representative color cannot be extracted in step S205, the process proceeds to step S209 and the matching degree is set to 0 and output.

In addition to the above, the positional relationship of each area may be taken into consideration. To compare the positional relationship of each area,
For example, the positional relationship of each color area can be compared by the position of the center of gravity of each area, or the position of the center point of a rectangle circumscribing the area can be compared.

Then, in step 104 of FIG.
If the matching degree is equal to or higher than a predetermined value, it is considered that the blocks existing in each area of FIG. 6 match the search image, and the inverse DCT calculation is performed on these blocks (step S105). If the matching degree is less than the predetermined value, the next frame is searched.

In step S106, step S105
The above-described matching process shown in FIG. 7 is performed on each block obtained by the inverse conversion in step S2, and the second-stage extraction process is performed.

As described above, in the second embodiment, the degree of matching is calculated by pattern matching. However, there are many methods other than those described here as the method of pattern matching, and any method may be used.

According to the second embodiment, for example, when an image showing a specific person is desired to be searched, one image showing the person is designated as a search image,
Since the person's area can be specified to some extent from the color of clothes, the area in which the inverse DCT calculation and the detailed matching are performed can be limited, and a high-speed and reliable search can be performed.

In the first and second embodiments, it may be selectable whether or not the extraction process by the second matching section 5 is performed. For example, uniform matching with DC component images can detect differences in uniforms, and fine matching after inverse DCT can detect differences in faces, uniform numbers, etc. Therefore, when it is desired to classify images by a team, it is possible to use DC component images. It is possible to perform only matching, and to perform detailed matching each time only when it is desired to classify the individual. In this way, matching processing can be selectively used as necessary, and the balance between search speed and certainty can be controlled.

As described above, according to the above-described embodiment, the first matching unit makes a rough judgment based on the average color of the blocks, restores only the blocks matching the conditions to the original image, and further The second matching unit makes a detailed determination based on the color of each pixel. Therefore, the inverse DC
It is possible to reduce the number of times T calculation and pattern matching for each pixel are performed, and it is possible to perform a search from a large number of frames at high speed.

Furthermore, by making it possible to switch between the search up to the first step by the first matching section and the search up to the second step by the second matching section, it is possible to perform the matching processing as necessary. There is an effect that it can be used properly and the balance between search speed and certainty can be controlled.

Further, the present invention may be applied to a system composed of a plurality of devices or to an apparatus composed of one device. Further, it goes without saying that the present invention can be applied to the case where it is achieved by supplying a program to a system or an apparatus. In this case, the storage medium storing the program according to the present invention constitutes the present invention. Then, by reading the program from the storage medium to the system or device, the system or device operates in a predetermined manner.

[0070]

As described above, according to the present invention,
Since the amount of processing at the time of image retrieval is reduced, it is possible to perform image retrieval from a large number of frames at high speed.

Further, according to the present invention, the number of times of performing inverse DCT calculation or pixel-based pattern matching at the time of image retrieval is reduced, and the image retrieval can be speeded up.

[0072]

[Brief description of drawings]

FIG. 1 is a block diagram showing an outline of a hardware configuration of an image search device according to the present embodiment.

FIG. 2 is a block diagram showing a functional configuration of the image search device according to the present embodiment.

FIG. 3 is a flowchart showing an operation procedure of the image search device according to the first embodiment.

FIG. 4 is a diagram showing a display for designating a color as a search condition.

FIG. 5 is a diagram illustrating an example of a search image according to the second embodiment.

FIG. 6 is a diagram illustrating an example of area division of a DC component image according to the second embodiment.

FIG. 7 is a flowchart showing a procedure of matching processing according to the second embodiment.

FIG. 8 is a diagram showing a block diagram of an image compression apparatus using an MPEG algorithm.

FIG. 9 is a diagram showing a general configuration of an apparatus for searching a frame that matches a desired condition from compressed data of a moving image.

FIG. 10 is a diagram showing an arrangement of frequency components by DCT conversion.

FIG. 11 is a diagram illustrating an example of a quantization table.

[Explanation of symbols]

 1 Buffer 2 Variable Length Decoding Section 3 Inverse Quantization Section 4 Inverse DCT Section 5 Second Matching Section 6 Search Condition Input Section 7 DC Component Separation Section 8 First Matching Section 9 Area Determining Section

Claims (12)

[Claims] 1. An image retrieving apparatus for retrieving a desired image from image data in which an image is divided into a plurality of blocks and each block is represented by a spatial frequency component, wherein a predetermined spatial frequency component of each block is obtained from the image data. Acquiring means for acquiring, a extracting means for extracting a block satisfying a preset search condition based on the predetermined frequency component acquired by the acquiring means, and a spatial frequency component of the block extracted by the extracting means. On the basis of the above, the restoring means for restoring the block to the image data by the bitmap, the determining means for determining whether or not the block restored by the restoring means satisfies the search condition, and the restoring means for restoring the block. And an output unit that outputs the image as a search result when it is determined that the block satisfies the search condition. Image retrieval apparatus. 2. The image retrieval apparatus according to claim 1, wherein the predetermined spatial frequency component acquired by the acquisition means is a direct current component. 3. The image search device according to claim 1, wherein the search condition is a color designated by an operator. 4. The display device further comprises display means for displaying a color sample and setting means for designating a desired color from the color sample displayed on the display means in order to set the search condition. The image search device according to claim 3. 5. When a block is extracted by the extraction means, the extracted image output means for outputting the image as a search result, and which of the extracted image output means and the output means outputs the search result is switched. The image search device according to claim 1, further comprising a switching unit. 6. An image retrieving apparatus for retrieving a desired image from image data in which an image is divided into a plurality of blocks and each block is represented by a spatial frequency component, wherein a predetermined spatial frequency component of each block is obtained from the image data. Acquisition means for acquiring, an extraction means for performing pattern matching with a preset search image based on the predetermined frequency component acquired by the acquisition means, and extracting a region that matches the search image, Based on the spatial frequency component of the area extracted by the extracting means, a restoring means for restoring the area to image data by a bitmap, and a pattern as to whether the area restored by the restoring means matches the search image When the determination unit that determines by matching and the determination unit determines that the restored area matches the search image, the image is searched for. Image retrieval apparatus according to an outputting means for outputting as. 7. The extraction means divides the search image into a plurality of color regions formed of colors in a predetermined similar range, and a predetermined division of the image formed of the predetermined spatial frequency component. It is characterized by further comprising: second dividing means for dividing into a plurality of color areas constituted by colors in the similar range; and comparing means for checking the matching by comparing the colors of the areas obtained by the first and second dividing means. The image search device according to claim 6. 8. The comparison means sets a region having the largest area of the plurality of color regions in the search image as a representative region, and the image composed of the predetermined spatial frequency components has the same similarity as the representative region. The image search device according to claim 7, wherein when an area having a range of colors is detected, the area is compared with another color area of the search image. 9. The comparing means further compares the ratio of the size of each color area in the search image to the ratio of the size of corresponding color areas of the image composed of the predetermined spatial frequency component. The image search device according to claim 8, characterized in that 10. When an area is extracted by the extraction means, the extracted image output means for outputting the image as a search result, and which of the extracted image output means and the output means outputs the search result is switched. The image retrieval apparatus according to claim 6, further comprising a switching unit. 11. An image retrieval method of dividing an image into a plurality of blocks and retrieving a desired image from image data expressing each block with a spatial frequency component, the predetermined spatial frequency component of each block from the image data. An extracting step of extracting a block satisfying a preset search condition based on the predetermined frequency component acquired in the acquiring step; and a spatial frequency component of the block extracted in the extracting step. A restoration step of restoring the block to image data by a bitmap based on the above, a determination step of determining whether or not the block restored in the restoration step satisfies the search condition, and a restoration step in the determination step. An output step of outputting the image as a search result when it is determined that the block satisfies the search condition. Image search method. 12. An image retrieval method for retrieving a desired image from image data in which an image is divided into a plurality of blocks and each block is represented by a spatial frequency component, wherein a predetermined spatial frequency component of each block is obtained from the image data. An acquisition step of acquiring, an extraction step of performing pattern matching with a preset search image based on the predetermined frequency component acquired in the acquisition step, and extracting a region that matches the search image, Based on the spatial frequency component of the area extracted in the extraction step, a restoration step of restoring the area to image data by a bitmap, and a pattern of whether the area restored in the restoration step matches the search image A determining step of determining by matching, and if the restored area is determined to match the search image in the determining step, the image is searched. Image retrieval method characterized by an output step of outputting as a result.