KR20110060454A - Apparatus and method for providing image identifying - Google Patents

Abstract

PURPOSE: An image identification apparatus and method are provided to offer the effectiveness of storage capacity and tenacity for geometrical distortion. CONSTITUTION: Using a specific variable and a user input signal, a resolution normalization unit(100) normalizes static images to a specific size. A feature area extractor(200) extracts an intrinsic specific area from the normalized image of the resolution normalization unit. Using the intrinsic specific area extracted through the feature area extractor, a global histogram generator(300) acquires an image identifier.

Description

Image identification device and method {APPARATUS AND METHOD FOR PROVIDING IMAGE IDENTIFYING}

The present invention relates to an image identification technology, and more particularly, to an image identification apparatus and method suitable for efficiently managing large-capacity images such as recognition of specific image information, search for identical images, and classification of similar images by extracting unique features from digital still images. will be.

The present invention is derived from a study conducted as part of the IT growth engine technology development project of the Ministry of Culture, Sports and Tourism [2007-S-017-03, Development of user-oriented content protection distribution technology].

Techniques for identifying deformed digital images have been continuously developed to date, and representative techniques in this field are divided into local and global methods.

The local method is the Scale Invariant Feature Transform (SIFT) method, and the edge method is the edge histogram method.

The local method has the advantage of being able to find partially similar images and being very robust against geometric attacks that change the axes, but has the disadvantage of generating a lot of performance degradation in terms of search time and storage.

In addition, the global method has a relatively large gain in search time and storage capacity, but has the disadvantage of being vulnerable to geometric attacks and having very low accuracy.

SUMMARY OF THE INVENTION The present invention has been made in view of the problems of the prior art, and is intended to enable fast data retrieval in a large capacity using a low capacity image identifier at a limited storage limit.

In addition, the present invention, by using the image identifier image distortion, for example, even if distortion, such as rotation, movement, cropping, brightness change, noise, format conversion, etc. of the image can be identified to delete the duplicate data and existing It is intended to realize fast and accurate image identification within limited time and limited storage capacity that cannot be solved by the methods.

According to an image identification apparatus for solving the problems of the present invention, a resolution normalization unit for normalizing a still image to a specific size and outputting a normalized image using a user input signal and a specific variable, and a normalized image from the resolution normalization unit. The apparatus may include a feature region extractor that extracts a unique feature region of an image as an input, and a global histogram generator that obtains an image identifier using the unique feature region extracted by the feature region extractor.

The user input signal may be a selection signal for determining whether to perform the normalization unit.

In addition, the specific variable may be width and height information of the still image.

The feature region extracting unit may include a feature point extracting unit extracting a feature point by detecting a corner point using a harris corner detection technique, and obtaining an intrinsic angle using ambient brightness information of the feature point, and extracting N invariant feature points. And a rotation property extractor to generate and a peripheral region extractor to generate a feature region based on the invariant feature points.

The feature points may be the top N feature points extracted in the order of intensity by the Harris corner detection technique.

The rotation attribute extractor may include a partial differential filter, a gradient histogram generator, an averaging processor, and a maximum angle processor.

The global histogram generator may include a feature region repeater, a sub-blocking feature generator, a brightness invariant factor extractor, a global histogram adder, and a wavelet processor.

According to an image identification method for solving the problems of the present invention, a process of outputting a normalized image by normalizing a still image to a specific size using a user input signal and a specific variable, and unique characteristics of the image by using the normalized image as an input And extracting an area, and obtaining an image identifier using the extracted unique feature area.

The user input signal may be a selection signal for determining whether to perform the normalization unit.

In addition, the specific variable may be width and height information of the still image.

The intrinsic feature region extracts feature points by detecting corner points using a Harris corner detection technique, generates N invariant feature points by obtaining an intrinsic angle using ambient brightness information of the extracted feature points, and generates the invariant feature points. It can be extracted by creating a feature region based on.

The feature points may be the top N feature points extracted in the order of intensity by the Harris corner detection technique.

According to the present invention, it is possible to provide the efficiency of storage capacity and the robustness against geometric distortion by extracting and using a fast search speed and a small image descriptor which are not solved by existing methods using local descriptors. In addition, it is possible to increase the search efficiency by providing a robust characteristic against the coordinate axis distortion that is not solved in the existing methods using the global descriptor.

Advantages and features of the present invention and methods for achieving them will be apparent with reference to the embodiments described below in detail with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, but can be implemented in various different forms, and only the embodiments make the disclosure of the present invention complete, and the general knowledge in the art to which the present invention belongs. It is provided to fully inform the person having the scope of the invention, which is defined only by the scope of the claims. Like numbers refer to like elements throughout.

In describing the embodiments of the present invention, if it is determined that a detailed description of a known function or configuration may unnecessarily obscure the gist of the present invention, the detailed description thereof will be omitted. In addition, terms to be described below are terms defined in consideration of functions in the embodiments of the present invention, which may vary according to intentions or customs of users and operators. Therefore, the definition should be based on the contents throughout this specification.

Combinations of each block of the accompanying block diagram and each step of the flowchart may be performed by computer program instructions. These computer program instructions may be mounted on a processor of a general purpose computer, special purpose computer, or other programmable data processing equipment such that instructions executed through the processor of the computer or other programmable data processing equipment may not be included in each block or flowchart of the block diagram. It will create means for performing the functions described in each step. These computer program instructions may be stored in a computer usable or computer readable memory that can be directed to a computer or other programmable data processing equipment to implement functionality in a particular manner, and thus the computer usable or computer readable memory. It is also possible for the instructions stored in to produce an article of manufacture containing instruction means for performing the functions described in each block or flowchart of each step of the block diagram. Computer program instructions may also be mounted on a computer or other programmable data processing equipment, such that a series of operating steps may be performed on the computer or other programmable data processing equipment to create a computer-implemented process to create a computer or other programmable data. Instructions that perform processing equipment may also provide steps for performing the functions described in each block of the block diagram and in each step of the flowchart.

In addition, each block or step may represent a portion of a module, segment or code that includes one or more executable instructions for executing a specified logical function (s). It should also be noted that in some alternative embodiments, the functions noted in the blocks or steps may occur out of order. For example, the two blocks or steps shown in succession may in fact be executed substantially concurrently or the blocks or steps may sometimes be performed in the reverse order, depending on the functionality involved.

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

FIG. 1 is a block diagram illustrating an image identification apparatus for extracting an image identifier according to an exemplary embodiment, and includes an input unit 1, a resolution normalization unit 100, a feature region extraction unit 200, and a global histogram generator 300. ) May be included.

The input unit 1 is a means for inputting a user selection signal for determining whether to perform the resolution normalization unit 100 described later. A specific variable, for example, the width W and the height H of the still image may be input through the input unit 1.

The resolution normalization unit 100 normalizes a still image to a specific size (W × H) using a user selection signal of the input unit 1 and a specific variable including a width (W) and a height (H). Can be obtained.

At this time, if the function is not performed under the input condition given to the input unit 1, the still image will be a normal image.

The feature region extractor 200 extracts a fixed number N of feature regions that may become a unique feature of the image by using the normal image as an input.

The global histogram generator 300 may acquire one image identifier that is robust to distortion by using the feature regions extracted by the feature region extractor 200.

FIG. 2 is a block diagram illustrating a feature region extractor 200 in the image identification apparatus of FIG. 1 in detail. The feature region extractor 200 of FIG. 2 includes a feature point extractor 210 and rotation attribute extraction. The unit 220 and the peripheral region extraction unit 230 may be included.

As shown in FIG. 2, the feature point extractor 210 extracts a feature point using a Harris Corner detection technique that detects a corner point, in order of intensity of the Harris corner detection technique from the feature point extractor 210. The information of the upper N feature points extracted as may be composed of coordinates (x, y) on the normal image, a size (s) from which the surrounding information is extracted, and a total of three values (x, y, s). In this case, the above-mentioned Harris corner can be easily understood by those of ordinary skill in the art, so a detailed technical description of the Harris corner will be omitted.

The rotation property extractor 220 obtains the intrinsic angle θ by using the peripheral brightness information of the feature point extracted from the feature point extractor 210, and sets a predetermined number, for example, N invariant feature points (x, y, s, θ).

The peripheral region extractor 230 may generate a feature region based on the invariant feature points. Here, the feature region may have a block size s × s based on the coordinates (x, y), and may include an intensity value of the rotated region using Equation 1 below. Thus, when the feature region is obtained after rotation, there is an advantage of having an invariant characteristic in deformation such as rotation.

FIG. 3 is a block diagram illustrating a detailed configuration of the rotation property extractor 220 of FIG. 2. The partial differential filter 22, the gradient histogram generator 24, the averaging processor 26, and the maximum angle processor 28 are illustrated in FIG. And the like.

Prior to the description of FIG. 3, the rotation property extractor 220 may calculate the property angle θ of a given feature region, which may be implemented by calculating an angle based on a gradient histogram, which is shown in FIG. Those skilled in the art will readily understand.

As illustrated in FIG. 3, the partial differential filter 22 performs partial differential filtering on the peripheral information of the normal image based on the x coordinate and the y coordinate of the feature points (x, y, s) to obtain two partial differential result values (Dx, Dy). ) Can be obtained.

The gradient histogram generator 24 may serve to generate a gradient histogram by calculating an intensity and an angle corresponding to the partial differential result values Dx and Dy.

The averaging processor 26 may play a role of performing noise reduction (averaging processing) of a set number, for example, K, on the gradient histogram generated through the gradient histogram generator 24.

Finally, the maximum angle processor 28 extracts the corresponding angle θ corresponding to the point having the maximum value from the noise-free gradient histogram provided from the averaging processor 26, and adds each attribute to the characteristic point to invariant feature points. It may serve to generate (x, y, s, θ).

4 is a block diagram illustrating a detailed configuration of the global histogram generator 300 of FIG. 1. The feature region repeater 302, the sub-blocking feature generator 304, and the brightness invariant factor extractor 306 are illustrated in FIG. , The global histogram adder 308, the wavelet processor 310, and the like.

As illustrated in FIG. 4, the feature region repeater 302 may read a predetermined number, for example, N feature regions, one by one, and deliver the feature to the next process.

The sub-blocking feature generator 304 divides any (one) feature region of the N feature regions into B × B sub-blocks, and corresponding subblocks (i, j) (see FIG. 5). Totally accumulate and store the pixels belonging to (S _{(i, j)} ), and give a unique block number of the corresponding sub-block (Equation 2). These two variables extracted from one feature region may be used as an input of the global histogram adder 308 described later.

block_id ( i , j ) = i + j * B

The brightness invariant factor extractor 306 may extract a minimum value I _min and a maximum value I _max from the pixel value, respectively, in order to make an image identifier strong in brightness in one feature region.

Here, the minimum value I _min is the lower αth value when the brightness values of the pixels are arranged in ascending order in the feature area, and the maximum value I _max may mean the upper αth value when the brightness values are arranged in ascending order. have.

The global histogram adder 308 uses the unique block number (block_id (i, j) = j + j * B) extracted from the sub-blocking feature generator 304 described above, and the brightness invariant factor extractor 306. The extracted minimum value (I _min ) / maximum value (I _max ) may be added to the corresponding histogram. The addition function of the global histogram adder 308 may be implemented through the following [Equation 3], [Equation 4] and [Equation 5].

The total number of histograms generated from the global histogram adder 308 using Equation 3, Equation 4, and Equation 5 is B × B equal to the number of sub-blocks divided. The size of each histogram may be defined as L.

The wavelet processor 310 may apply resource reduction to a plurality of histograms generated by the global histogram adder 308 described above. Resource reduction can be applied as needed for fast calculation and small capacity.

Here, when the wavelet processor 310 applies the dimension reduction, the wavelet processor 310 may take a coefficient of a low part (eg, a low frequency region).

WS inputted to the wavelet processor 310 is a wavelet selector, and selects one of various wavelet filters (for example, a Harr, 5/3 taps, 9/7 taps filter, etc.) or does not wavelet. (Eg, no operation mode, "nothing"). WL represents the number of wavelet filter processes.

The global histograms finally generated through the above-described configuration and processing may be defined as unique image identifiers of still images in the present embodiment.

On the other hand, the method of calculating the similarity between the two images by using the image identifier in the present embodiment is as shown in Equation 6.

In addition, although the accuracy is slightly lower than [Equation 6], the similarity may be calculated by [Equation 7] or [Equation 8] for a large-capacity high-speed search.

As described above, in the present embodiment, a small image identifier enables fast data retrieval in a large capacity at a limited storage limit, and a slight image distortion, for example, rotation and movement of the image, is performed using the image identifier. The image can be identified even when distortion occurs such as clipping, brightness change, noise, and format conversion, so that duplicate data can be deleted and identified quickly and accurately within limited storage capacity and time that cannot be solved by conventional methods.

1 is a block diagram of a system for extracting an image identifier of a still image according to an exemplary embodiment of the present invention;

FIG. 2 is a detailed block diagram of the feature region extractor shown in FIG. 1;

3 is a detailed block diagram of the rotational constant extractor shown in FIG.

4 is a detailed block diagram of the global histogram generator of FIG. 1;

5 is an exemplary diagram of sub-blocking one feature region into B × B.

<Brief description of the main parts of the drawing>

100: resolution normalization unit

200: feature region extraction unit

300: global histogram generator

210: feature point extraction unit

220: rotation property extractor

230: peripheral region extraction unit

302: feature region repeats

304: sub-blocking feature generator

306: brightness invariant extraction unit

308: global histogram adder

310: wavelet processing unit

Claims (1)

A resolution normalizer for outputting a normalized image by normalizing a still image to a specific size using a user input signal and a specific variable; A feature region extraction unit configured to extract a unique feature region of the image by inputting a normalized image from the resolution normalization unit; A global histogram generator that obtains an image identifier using the unique feature region extracted by the feature region extractor Image identification apparatus comprising a.

Images