KR100653026B1 - Apparatus and method for retrieving an arbitrary shaped texture - Google Patents

Abstract

The present invention relates to a method and apparatus for searching for a texture having an arbitrary shape in a texture image search. In particular, when an image is composed of pieces of a plurality of textures, the present invention expresses the texture characteristics of the image and the texture of each object The present invention relates to a texture retrieval method and apparatus for retrieving an object having the same texture regardless of its shape after obtaining a feature value.

Multimedia search, texture feature, MPEG-4, MPEG-7, radon feature, heirloom feature, wavelet feature, low nick feature.

Description

FIELD AND METHOD FOR RETRIEVING AN ARBITRARY SHAPED TEXTURE}             

1 is a flow chart of one embodiment of any texture retrieval method of the present invention;

2 is a block diagram of one embodiment of any texture retrieval apparatus of the present invention;

3 is a flow chart of another embodiment of any texture retrieval method of the present invention;

4 is a block diagram of another embodiment of any texture retrieval apparatus of the present invention.

The present invention relates to a method and apparatus for searching for a texture of any shape. More particularly, when an image is composed of a plurality of pieces of texture, the texture feature of the image is expressed and the texture feature value of each object is obtained. The present invention relates to a texture retrieval method and apparatus for retrieving an object having the same texture regardless of its shape.

To date, data retrieval has primarily been a language-based search. However, in recent years, due to the development of multimedia and the development of the Internet, a large amount of multimedia data exists on the Internet or in a multimedia database, so that not only text, but also movies, composite images, still images, audio, video, music, etc. The necessity of multimedia data retrieval, which is made up of and the like, has emerged. Accordingly, there is an increasing demand for an effective retrieval method for easily retrieving multimedia data that a user wants to search from a large data such as the Internet or a multimedia database.

Unlike existing data consisting of text only, multimedia data is much larger than text data and contains various types of information such as video, sound, and text. Therefore, multimedia data can be used to search for desired multimedia data. It is almost impossible. Therefore, in order to search for multimedia data, when constructing a multimedia database, a search must be performed by extracting a feature capable of expressing each multimedia data through a preprocessing process and comparing the extracted features with each other.

For the effective retrieval of such data, it is important to consider what properties of multimedia data are used, how to express the features, how to extract the features, and how to compare them. Here, a specific method of expressing the characteristics of the multimedia data is called a descriptor. In multimedia data, video and audio can be considered as the most important parts. In the case of video, the features such as color, shape, texture, and motion are extracted and similarity is extracted. Will be compared. For example, color histogram, correlogram, etc. can be used as descriptors to express colors.In the case of shapes, the boundary can be expressed by direction code and searched or various transformations can be used. It can be expressed and searched regardless of slight deformation, movement, rotation, and scale. For motion, similar video sample sequences can be retrieved in the same way as a histogram of motion in a video.

In theory, all images are composed of textures, so texture retrieval is a very important part of image retrieval. Searching for the texture of an image is to find an image with a similar pattern (texture) as a person looks at the pattern of another tree and thinks it as the same pattern.

In the recent case of MPEG-4, instead of encoding and transmitting the entire frame, object-based encoding that encodes and transmits or manipulates only a specific region or object that a user is interested in or needs. An object-based coding scheme has been proposed and actively studied. In such object unit encoding, a part or region to be encoded is called an object, and a part or region which is not an object is called a background. When the object-based coding scheme is applied to MPEG-7 image retrieval, a specific object region may be described as a texture, a color, or a shape which is a descriptor of MPEG-7. In this case, it is not a big problem to handle it in the case of color or shape, but in the case of a texture, it may be difficult to extract stable texture features if the shape is different. Therefore, there is an urgent need to develop a texture retrieval method capable of classifying and retrieving objects having the same texture regardless of shape.

An object of the present invention is to overcome the problems of the prior art described above, after extracting and storing a texture feature value from a texture consisting of any shape, and then using the above to mask the shape of the comparison target to the reference texture to search The present invention provides a texture information retrieval method and apparatus for classifying and retrieving objects having the same texture regardless of shape, by extracting and comparing texture feature values in the same manner.

That is, one aspect of the present invention

Masking a texture to be searched by the user for the shape of the comparison target by receiving shape information of the query image (S1);

A texture feature extraction step (S2) of extracting feature values of a texture to be searched;

Another feature extraction step (S3) of extracting texture features of respective image data in the multimedia database by the same process as the texture feature extraction step;

Database building step (S4) for building a database with the feature-extracted multimedia data:

A similarity comparison step (S5) of comparing the texture feature of the output query image with the texture feature of the images in the database; And                         

And a comparison result outputting step (S6) of outputting the similarity comparison result of the previous step.

Another aspect of the present invention comprises the step of putting a texture of any shape into a rectangle and filling the empty portion of the texture with the same texture (S11);

A texture feature extraction step (S12) of extracting feature values of the texture in the square of the previous step;

Database construction step (S13) to database the texture features obtained in the previous step (S13):

Selecting a texture to be searched by a user and putting a texture to be searched into a rectangle and filling a blank portion of the texture with the same texture (S14);

Another texture feature extraction step (S15) of extracting a texture feature in the rectangle in a database by the same process as the texture feature extraction step (S12) following the previous step;

A similarity comparison step of comparing the texture feature of the output query image with the texture feature of the images in the database (S16); And

And a search result for a texture of any shape, comprising a comparison result output step (S17) for outputting the similarity comparison result of the previous step.

Hereinafter, the present invention will be described in more detail with reference to the accompanying drawings.

The present invention relates to a method and apparatus for searching for a texture of any shape that can search or classify textures regardless of such shapes, assuming that a general natural image is composed of a plurality of objects and each object has a different texture. will be.

1 is a flow chart of one embodiment of any texture retrieval method of the present invention. When searching for multimedia data by the texture retrieval apparatus of the present invention, first, the shape information of the query image is input to mask the texture to be searched by the shape of the object to be compared (S1). As used herein, shape information of an object means information for distinguishing which pixel belongs to an object and which pixel belongs to a background in an image.

Subsequently, the feature value of the texture to be searched is extracted. In this case, the feature extraction step may be performed by putting a texture of any shape into a minimum boundary rectangle including the shape and extracting a texture feature (S2).

In the database building step (S3), the texture feature of each image data in the database is extracted by the same process as the texture feature extraction step (S4). Then, comparing the similarity between the texture feature of the output query image and the texture feature of the images in the database (S5); Finally, the similarity comparison result of the previous step is output (S6).

In the feature extraction step (S2 and S4) of the present invention, a texture of any shape is put into a rectangle, and a Gabor feature value, a Radon feature value, a Wavelet feature value, or Zernike One of the feature values can be extracted. Gabor and Wavelet feature values are used to decompose the frequency of the texture. For example, Gabor filters are most commonly used because they can divide the frequency domain most effectively. The filter is widely used because it can divide the frequency domain into shapes that consider human visual characteristics. The Zernike feature value describes the feature of the image by calculating the rotational constant Gernique moment extracted from the normalized DFT image. Radon feature values are used to convert a two-dimensional image signal into a one-dimensional signal (sinogram) and then decompose the frequency of the sinogram by "central slice theorem."

The texture retrieval apparatus for implementing one embodiment of the texture retrieval method of the present invention, as shown in Figure 2, receives the shape information of the query image to mask the texture to the user to search the shape of the comparison target Masking unit 10; A texture feature extracting unit 20 for extracting feature values of a texture in the masked shape; Another feature extracting unit 20 'which extracts texture features of respective image data in the multimedia database by the same operation as that of the texture feature extracting unit; The multimedia database construction unit 30 constructing a multimedia database from the extracted feature multimedia data: Similarity comparing the texture feature of the query image output by the feature extractor of the query image with the texture feature of the images in the multimedia database. Comparator 50; And a comparison result output unit 60 which receives and outputs the result of the similarity comparison unit of the front end. In the texture retrieval apparatus of the present invention, the feature extractor 20 or 20 'extracts one of a Gabor feature value, a radon feature value, a wavelet feature value, or a low nick feature value by putting an arbitrary texture into a rectangle. It may be implemented as a device.

Referring to FIG. 3, in another embodiment of the texture retrieval method of the present invention, when a texture retrieval is performed, a texture of an arbitrary shape is inserted into a rectangle, and a blank portion of the texture is filled with the same texture (S11). The feature values are extracted (S12), and these texture features are databased (S13).

On the other hand, after the user selects the texture of the query image to be searched, the texture to be searched is put in a rectangle and the empty portion of the texture is filled with the same texture (S14). The texture feature in the rectangle is extracted by the same process as the texture feature extraction step (S12) (S15). In the present invention, the setting of the rectangle may be limited to a minimum bounding rectangle surrounding an arbitrary shape, that is, a tightest rectangle having the smallest inner area while including an object therein. Subsequently, the similarity between the query image and the images in the database is compared based on the texture feature of the query image obtained in steps S12 and S15 (S16), and the image having the closest texture is found as a result of the similarity comparison. Outputs (S17).

An example of an apparatus for implementing such a search method is shown in FIG. 4. The texture retrieval apparatus of the present invention includes a unit 110 for putting a texture of an arbitrary shape of a query image to be searched by a user into a rectangle and filling a blank portion with the same texture; A texture feature extractor 120 for extracting feature values of the texture; A unit (110 ') which receives texture data and puts a texture of any shape into a rectangle and fills an empty portion of the texture with the same texture; Another feature extractor (120 ′) extracting texture features of respective image data in the multimedia database by the same operation as that of the texture feature extractor (120); The multimedia database constructing unit 130 constructing a multimedia database using the extracted feature multimedia data: Similarity comparing the texture feature of the query image output by the feature extractor of the query image with the texture feature of the images in the multimedia database. Comparator 140; And a comparison result output unit 150 that receives and outputs the result of the similarity comparison unit of the front end.

In the second embodiment of the present invention, the feature extraction may be implemented by extracting any type of texture into a rectangle and extracting one of Gabor, Radon, Wavelet, or Lowk.

The texture retrieval method and apparatus of the present invention express an image when an arbitrary image is composed of several pieces of texture, obtain texture feature values on each object, and then search for an effective texture retrieval that can search for objects having the same texture regardless of shape. Provide means.

Claims (18)

Masking a texture to be searched by the user for the shape of the comparison target by receiving shape information of the query image (S1); A texture feature extraction step (S2) of extracting feature values of a texture to be searched; Another feature extraction step (S3) of extracting texture features of respective image data in the multimedia database by the same process as the texture feature extraction step; Database building step (S4) for building a database with the feature-extracted multimedia data: A similarity comparison step (S5) of comparing the texture feature of the output query image with the texture feature of the images in the database; And And a comparison result output step (S6) of outputting the similarity comparison result of the previous step. 2. The texture retrieval of claim 1, wherein the step of extracting features comprises extracting a Gabor texture feature by placing a texture of any shape into a minimum bounding rectangle around the shape of the texture. Way. 2. The texture retrieval of any of claims 1 to 6, wherein the step of extracting the features is a step of putting a texture of any shape into a minimum bounding rectangle around the shape of the texture and extracting radon feature values. Way. 2. The texture search of any of claims 1 to 6, wherein the step of extracting features comprises extracting a texture of any shape into a minimum bounding rectangle around the shape of the texture and extracting wavelet feature values. Way. 2. The texture of any shape according to claim 1, wherein the feature extraction step is to extract a texture of any shape into a minimum bounding rectangle around the shape of the texture and extract a Zernike feature value. Search method. A masking unit which receives shape information of a query image and masks a texture to be searched by a user into a shape of a comparison target; A texture feature extracting unit for extracting feature values of a texture in the masked shape; Another feature extractor which extracts texture features of respective image data in a multimedia database by the same operation as that of the texture feature extractor; A multimedia database construction unit for constructing a multimedia database with the extracted feature multimedia data: A similarity comparison unit for comparing the feature of the query image output by the feature extractor of the query image with the features of the images in the multimedia database; And And a comparison result output unit configured to receive and output a result of the similarity comparison unit of a previous stage. 7. The apparatus of claim 6, wherein the feature extractor is means for extracting Gabor feature values by placing a texture of any shape into a minimum bounding rectangle around the shape of the texture. . 7. The apparatus of claim 6, wherein the feature extractor is means for extracting a Radon feature value by placing a texture of any shape into a minimum bounding rectangle around the shape of the texture. . 7. The method of claim 6, wherein the feature extractor is means for extracting a wavelet feature value or a lowk feature value by placing a texture of any shape into a minimum bounding rectangle around the shape of the texture. Shape texture search device. Inserting a texture of an arbitrary shape into a rectangle and filling a blank portion with the same texture (S11); A texture feature extraction step (S12) of extracting feature values of the texture in the square of the previous step; Database construction step (S13) to database the texture features obtained in the previous step (S13): Selecting a texture to be searched by a user and putting a texture to be searched into a rectangle and filling a blank portion of the texture with the same texture (S14); Another texture feature extraction step (S15) of extracting a texture feature in the rectangle in a database by the same process as the texture feature extraction step (S12) following the previous step; A similarity comparison step of comparing the texture feature of the output query image with the texture feature of the images in the database (S16); And And a comparison result output step (S17) of outputting the similarity comparison result of the previous step. 11. The method of claim 10, wherein extracting the feature is extracting Gabor texture features. 11. The method of claim 10, wherein the feature extraction step is a step of extracting radon feature values. 11. The method of claim 10, wherein extracting the feature is extracting a wavelet feature value. 11. The method of claim 10, wherein the feature extraction step is a step of extracting Zernike feature values. A unit for inserting a texture of an arbitrary shape to be searched by a user into a rectangle and filling an empty portion of the texture with the same texture; A texture feature extraction unit for extracting feature values of the texture; A unit for inputting texture data and putting a texture of any shape into a rectangle and filling an empty portion with the same texture; Another feature extractor which extracts texture features of respective image data in a multimedia database by the same operation as that of the texture feature extractor; The multimedia database construction unit for constructing a multimedia database from the extracted feature multimedia data: a similarity comparison unit for comparing the texture features of the query image output by the feature extraction unit of the query image and the texture features of the images in the multimedia database; And a comparison result output unit configured to receive and output a result of the similarity comparison unit of a previous stage. 16. The apparatus of any one of claims 15 to 17, wherein said feature extractor is means for extracting Gabor feature values. 16. The apparatus of any one of claims 15 to 17, wherein said feature extractor is a means for extracting radon feature values. 16. The apparatus of claim 15, wherein the feature extractor is a means for extracting wavelet feature values or low nick feature values.

Images