KR20180054406A - Image processing apparatus and method - Google Patents

Abstract

An image processing device comprises: a calculation part calculating a location offset related to each of a plurality of candidate selection regions in a second frame based on a location of a basis image in a first frame; and a determination part determining a final selection region including a target in the second frame according to the calculated location offset and a weight assigned to each of the plurality of candidate selection regions.

Description

[0001] IMAGE PROCESSING APPARATUS AND METHOD [0002]

The following description relates to image processing technology.

Portable terminals such as a smart phone, a personal computer (PC), a laptop computer, and the like are becoming popular. Thus, a target of interest can be continuously photographed through a photographing apparatus in the portable terminal. In order to acquire a clear frame in a video file continuously shot, it is necessary to perform a process of tracking the target and adjusting the focus accordingly.

The target tracking method includes, for example, a method of performing selective search for a specific bounding box among a plurality of regions, a method of using a particle filter for predicting stochastic movement of each point, And tracking the target in the frame.

According to one aspect, there is provided a computing unit for calculating a position offset for each of a plurality of candidate selection regions in a second frame based on a position of a basis image in a first frame, And a determination unit that determines a final selection area including the target in the second frame according to a weight assigned to each of the areas.

According to one embodiment, the determination unit may determine the weight based on a position in the second frame of each of the plurality of candidate selection areas.

According to another embodiment, the calculation unit calculates a plurality of position offsets by applying a characteristic regression matrix to each of the plurality of candidate selection regions, and applies the weight to each of the plurality of position offsets to track the target The target position offset can be calculated.

According to another embodiment, the calculation unit may calculate a plurality of position offsets with respect to the first candidate selection region using a plurality of previously learned feature regression matrices, and calculate a plurality of position offsets using the average of the plurality of position offsets, And calculate a first position offset corresponding to the candidate selection area. Further, the plurality of feature regression matrices may be learned based on feature points in the base image and position offsets according to feature points of each of the plurality of sample frames.

According to another embodiment, the image processing apparatus may determine an initial selection region related to the target in the second frame based on a base image in the first frame, and determine, based on the determined initial selection region, And an extraction unit for extracting the candidate selection area. The extracting unit may calculate an overall position offset between the first frame and the second frame, and may determine an initial selection area based on the calculated total position offset and position information of the target in the base image.

According to another embodiment, the extraction unit extracts projection points corresponding to feature points in the base image in the second frame, and uses texture values of the extracted projection points and points within a predetermined range To determine the overall position offset. Wherein the extraction unit extracts points within a predetermined range around the extracted projection point, determines a matching point corresponding to the feature point according to a similarity between a texture value of the extracted points and a texture value of the feature point, The position of the point and the position of the matching point may be compared to determine the overall position offset.

According to another embodiment, the image processing apparatus stores the second frame in which the final selection area is determined, and when the number of stored frames becomes equal to or greater than a threshold value, And a storage unit for updating the storage unit.

According to another aspect, an image processing method includes calculating a similarity of each of a positive sample and a plurality of candidate selection regions with respect to a characteristic of a target, and determining a final selection And determining a region.

According to an exemplary embodiment, the step of calculating the similarity may include comparing the features of the positive sample included in the Sparse Subspace Clustering (SSC) model with the features of the sub-regions within each candidate selection region, And a step of calculating.

According to another embodiment, calculating the similarity may include calculating the similarity of the first candidate selection region according to the sum of the similarities with respect to the plurality of subareas included in the first candidate selection region .

According to yet another embodiment, the sparse subspace clustering model can be learned from a plurality of sample frames based on the Euclidean distance of the positive sample with respect to the characteristic of the target and the negative sample with respect to the characteristic of the neighborhood of the target have.

According to another embodiment of the present invention, the image processing method further comprises the steps of: comparing a similarity degree between the last selected region of the frame and the positive sample, a similarity average value between the previous frames and the positive sample; And storing it. More specifically, the image processing method may further comprise the step of comparing the number of stored frames with a predetermined threshold, and updating the sparse subspace clustering model using the stored frames as a sample frame according to a result of the comparison have.

According to another aspect of the present invention, there is provided an apparatus for calculating a target position offset for tracking a target from feature points existing in each of a plurality of candidate selection regions, A second calculator for calculating a similarity of each of the plurality of candidate selection regions with respect to a positive sample with respect to a feature of the target, and a second calculation unit for applying a first weight to the target position offset, And a determination unit for determining a final selection region including the target by applying a second weight to the target.

According to one embodiment, the first calculator calculates a plurality of position offsets by applying a characteristic regression matrix to each of the plurality of candidate selection regions, and applies a weight to the calculated plurality of position offsets to calculate the target position offset Can be calculated.

According to another embodiment, the second calculation unit may calculate a target value of the target according to a Hybrid Sparse Subspace Clustering (HSSC) model learned using a positive sample of the target and a negative sample of a neighboring region of the target. The degree of similarity can be calculated.

According to another aspect, there is provided a target tracking method comprising obtaining a candidate selection region for a target of a current frame and proceeding with characteristic regression for the obtained candidate selection region to obtain a final selection region.

According to one embodiment, the step of acquiring the candidate selection region relating to the target among the current frames includes the steps of: determining information of an initially selected region of the target of the current frame based on a base image of the target among the previously stored frames; And obtaining information of a first set number of candidate selection areas around the initial selection area of the target.

According to another embodiment, the step of performing the feature regression for the obtained candidate selection region to obtain the final selection region includes the steps of: regressing the information of each candidate selection region obtained based on the feature regression matrix; And determining a final selected region of the target based on information after comprehensive regression of the candidate selection region obtained by the regression.

According to still another embodiment of the present invention, the step of performing the feature regression for the obtained candidate selection region to acquire the final selection region includes the steps of: performing the feature regression and the feature evaluation on the obtained candidate selection region, The method comprising the steps of:

According to another aspect, the method includes obtaining a candidate selection region with respect to a target of a current frame, and performing a feature evaluation on the obtained candidate selection region to obtain a final selection region, A progress tracking method based on the clustering model is provided.

According to an exemplary embodiment, the step of performing feature evaluation on the obtained candidate selection region to obtain a final selection region includes evaluating information of each candidate selection region acquired based on the sparse partial space clustering model And determining the final selected region of the target based on the information of the candidate selection region of the maximum evaluation value obtained by the evaluation.

According to another embodiment, the step of evaluating information of each candidate selection area acquired based on the sparse subspace clustering model may include the step of evaluating information of each candidate selection area obtained based on the sparse subspace clustering model Determining an evaluation value of an image feature of the candidate selection area based on evaluation values of image features of each sub-candidate selection area; And determining the maximum evaluation value from the evaluation value of the area feature.

FIG. 1 is a diagram illustrating an operation of an image processing apparatus according to an exemplary embodiment of the present invention.
2 is a block diagram illustrating an image processing apparatus according to an embodiment.
FIG. 3A and FIG. 3B are views illustrating a process of determining an initial selection region using a base image by the image processing apparatus illustrated in FIG. 2. FIG.
4 is a flowchart illustrating a process of determining a final selection area including a target by the image processing apparatus illustrated in FIG.
FIG. 5 is a flowchart illustrating a process of learning a characteristic regression matrix by an image processing apparatus according to another embodiment.
FIG. 6 is a flowchart illustrating a process of determining a final selection area including a target according to another embodiment of the present invention.
FIGS. 7A and 7B are diagrams for explaining a process of extracting a positive sample and a negative sample using a sample frame according to another embodiment of the present invention.
FIG. 8 is a flowchart illustrating a process of learning a sparse partial space clustering model by the image processing apparatus illustrated in FIG. 7A.
9 is a block diagram illustrating an image processing apparatus according to another embodiment.

It is to be understood that the specific structural or functional descriptions disclosed herein may be implemented by way of example only, and that the embodiments may be embodied in various other forms and are not limited to the embodiments described herein It does not.

The terms first or second may be used to describe various components, but such terms should be understood only for the purpose of distinguishing one component from another. For example, the first component may be referred to as a second component, and similarly, the second component may also be referred to as a first component.

The singular expressions include plural expressions unless the context clearly dictates otherwise. As used herein, the terms "comprise", "comprising" and "comprising" are used herein to designate the presence of stated features, integers, steps, operations, elements, components, or combinations thereof, Elements or parts thereof, or combinations thereof, without departing from the spirit and scope of the invention.

Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art. Terms such as those defined in commonly used dictionaries are to be interpreted as having a meaning consistent with the meaning of the context in the relevant art and, unless explicitly defined herein, are to be interpreted as ideal or overly formal Do not.

Hereinafter, embodiments will be described in detail with reference to the accompanying drawings. Like reference symbols in the drawings denote like elements.

FIG. 1 is a diagram illustrating an operation of an image processing apparatus according to an exemplary embodiment of the present invention. The image processing apparatus 100 may be implemented as a computer device. For example, the image processing apparatus 100 may be implemented with at least one software module, at least one hardware module, or various combinations thereof. Referring to FIG. 1, the image processing apparatus 100 may receive an input image and determine a final selection area including a target in the received input image. In the following description, the target means an object that is tracked in the image data as a region of interest of the user. For example, the target may be designated as a particular user's eye and nose within the frame of the entire image. Alternatively, the target may be designated as the position of a particular player within a sports image in which a plurality of players move together.

In order to determine the final selection area containing the target, the image processing apparatus 100 may use a pre-stored basis image. In the following description, the base image means an area determined as a target in a specific frame. The base image may be stored in various forms such as a polymorphic region, an elliptical region, and an irregular region connecting a plurality of pixel points. Illustratively, polymorphism can be implemented in the form of triangles, pentagons, pentagons, hexagons, and the like.

More specifically, the image processing apparatus 100 can determine the final selection area 110 in the first frame F ₁ based on the position of the base image. Similarly, the image processing apparatus 100 can determine the final selection area 120 in the Nth frame F _N based on the position of the base image. The image processing apparatus 100 may update the base image according to the number of frames for which the final selection area is determined. The process of selecting and updating the base image will be described in more detail below.

2 is a block diagram illustrating an image processing apparatus according to an embodiment. 2, the image processing apparatus 200 may include an extraction unit 210, a calculation unit 220, a determination unit 230, and a storage unit 240. The extracting unit 210 can determine an initial selection area for the target in the second frame based on the base image in the first frame. Illustratively, the first frame is an arbitrary frame existing in the entire input image, and may indicate a frame in which information related to the base image is stored in advance. The base image may represent an image of a region including a target in the first frame. Illustratively, the information about the base image may represent information about feature points existing in the base image. In addition, the second frame may indicate a frame input to the image processing apparatus 200 after the first frame in the entire input image.

The extracting unit 210 may extract a plurality of candidate selection regions around the initial selection region determined in the second frame. Illustratively, the extraction unit 210 may extract a predetermined number of candidate selection regions around the initial selection region with respect to the target.

The calculation unit 220 may calculate a position offset of each of the plurality of candidate selection regions in the second frame based on the position of the base image in the first frame. In one embodiment, the calculator 220 may perform a feature regression for each of the plurality of candidate selection regions to calculate the position offset. In the following description, feature regression refers to the process of tracking a position offset in which feature points in a base image are shifted for each candidate selection region.

In one embodiment, the calculation unit 220 may calculate a plurality of position offsets by applying a learned feature regression matrix to each of the plurality of candidate selection regions. In the following description, the feature regression matrix may represent a determinant that defines the positional difference between the feature points in the base image and the feature points in the candidate selection area. The image processing apparatus 200 compares the position of the base image in which the target exists and the size and position of the candidate selection region on the basis of the feature regression matrix and selects candidates having the most similar characteristics to the base image, You can track areas.

In addition, the calculation unit 220 may calculate a target position offset that tracks the target by applying a weight to each of the calculated plurality of position offsets.

The determination unit 230 may determine a final selection area including the target in the second frame according to the calculated position offset and the weight assigned to each of the plurality of candidate selection areas. The determination unit 230 may determine the weight according to the position in which each of the plurality of candidate selection regions is present in the second frame.

The storage unit 240 may store the second frame for which the final selection area is determined in the memory in the image processing apparatus 200. [ In addition, when the number of stored frames is equal to or greater than the threshold value, the storage unit 240 may update the base image according to the target tracking result of the frames.

In another embodiment, the storage unit 240 may update the feature regression matrix using the newly stored frame. In addition, the storage unit 240 may replace the least recently updated feature regression matrix as a new feature regression matrix.

FIG. 3A and FIG. 3B are views illustrating a process of determining an initial selection region using a base image by the image processing apparatus illustrated in FIG. 2. FIG. Referring to FIG. 3A, the image processing apparatus may determine an initial selection region 340 in a second frame F _{2 that} is newly input using a first frame F _{1 that} is stored in advance. In one embodiment, the extractor 210 may determine a plurality of matching points 330 in the second frame F ₂ based on feature points 320 in the base image 310. In the following description, the feature point means a point among the pixel points in the frame that can be identified by the image processing apparatus. For example, the feature point may be defined as points of the face contour point of the target and points such as a face organ (eye, nose, mouth, etc.) point.

The extracting unit 210 may extract a plurality of projection points in the second frame F ₂ of the input image using the feature points 320 in the first frame F ₁ stored in advance. In the following description, the projection point may represent a pixel point in a second frame corresponding to the feature points 320. [ The extraction unit 210 may extract pixel points existing within a predetermined range around the projection point. Illustratively, the extraction unit 210 may extract the points corresponding to the 3-by-3 matrix about the position of the projection point as the pixel points. The size of the third row and third column should be construed as illustrative only for the sake of understanding, and should not be construed as limiting or limiting the scope of other embodiments, and a wide range of pixel points may be extracted according to the choice of an expert in the art.

As an example, the extracting unit 210 may compare the texture value of the feature points 320 with the texture value of the extracted pixel points, and determine a matching point in the second frame according to a result of the comparison . Illustratively, the extractor 210 may determine pixel points having the highest similarity to the texture values of the feature points 320 as matching points in the second frame.

In another embodiment, the extracting unit 210 calculates a texture gradient value of the projection point and the extracted pixel points, and extracts pixel points whose calculated texture gradient value exceeds a predetermined threshold value as a candidate matching point can do. The extracting unit 210 may compare the texture value of the feature points 320 with the texture value of the candidate matching points and determine the matching point 330 in the second frame according to the result of the comparison. More specifically, the extracting unit 210 may determine the pixel points having the highest similarity to the texture value of the feature points 320 as the matching point 330 in the second frame. The image processing apparatus according to the present embodiment can select the initial selection area for tracking the target based on the projection point, thereby improving the accuracy and efficiency of the target tracking.

In one embodiment, the extractor 210 compares the position of the feature points 320 in the first frame F ₁ and the position of the match points 330 in the second frame F ₂ And determine an overall position offset for the second frame (F ₂ ) of the input image according to the result of the comparison.

In another embodiment, the extractor 210 may use the weights corresponding to the respective feature points 320 to determine the overall position offset for the second frame (F ₂ ) of the input image. More specifically, the extracting unit 210 calculates a position offset between each feature point 320 and the matching point 330, and outputs a weighting average value obtained by applying the weight to the calculated position offset, Can be calculated as the total position offset with respect to the two frames (F ₂ ). Illustratively, the weight may be determined according to the texture value of the feature point 320 and the similarity of the texture value of the match point 330.

The extracting unit 210 extracts region information of the initial selection region 340 in the second frame F ₂ using the region information of the base image 310 in the first frame F ₁ and the calculated total position offset Can be calculated. Illustratively, the area information may include at least one of image data in the area, size information of the area, position information of the area within the frame, and information about the feature points in the area.

Extraction unit 210 includes a first frame (F ₁₎ located in the base image (310) in the information and a second frame (F ₂₎ on the basis of the total offset position with the second frame (F ₂₎ target within about Can be estimated. The position information of the target may indicate a position where the base image 310 exists in the second frame F ₂ . In addition, the extracting unit 210 may determine the initial selection region 340 about the target in the second frame F ₂ according to the estimated position information of the target. Illustratively, the size of the initial selection region 340 and the size of the base image 310 may be the same.

Referring to FIG. 3B, the image processing apparatus can extract a plurality of candidate selection regions 351, 352, 353, and 354 based on the extracted initial selection region 340. In one embodiment, the extractor 210 samples the candidate selection area X _i (where i = 1, ..., N, N is a positive integer) within the neighboring space of the initial selection area 340, It is possible to obtain the area information on the candidate selection area X _i . The region information may include size information of the candidate selection region X _i , position information, and information of feature points 331, 332, 333, 334, 335, and 336 in the region.

As another example, if the base image includes at least two or more base sub-images, the extracting unit 210 obtains the information of the sub-candidate selection regions corresponding to the respective base sub-images based on the extracted candidate selection region can do. As an example, there may be a case where one base image is divided into four base sub images and stored in the memory. In this case, the extracting unit 210 can extract the sub candidate selection regions respectively corresponding to the four base sub images and obtain information corresponding to each of the sub candidate selection regions.

4 is a flowchart illustrating a process of determining a final selection area including a target by the image processing apparatus illustrated in FIG. Referring to FIG. 4, the image processing apparatus includes a step 410 of calculating a plurality of position offsets by applying a characteristic regression matrix to each of a plurality of candidate selection regions, applying a weight to each of the plurality of position offsets, (Step 420), and determining (430) a final selection area containing the target using the target position offset.

In step 410, the image processing apparatus may perform feature regression to calculate the difference between the position of the base image in the first frame and the representative position of each of the candidate selection regions in the second frame. More specifically, the image processing apparatus can calculate a plurality of position offsets by applying a learned feature regression matrix to each candidate selection region.

There may be a case where the representative position of the candidate selection area is (x, y). In this case, the feature regression matrix H may be defined as a matrix H = [h ₁ , h ₂ ], which causes the representative position (x, y) to be featured within the base image. The feature regression matrix may represent a matrix that is learned based on feature points in the base image and feature points in the sample frame. A more detailed description of the process by which a regression matrix is generated mechanically is described in more detail below with reference to the figures to be added.

According to one embodiment, the image processing apparatus calculates, based on the feature point q _i existing in the candidate selection area X _i (where i = 1, ..., N, N is a positive integer) The position offset T _i corresponding to the candidate selection area X _i can be calculated.

In Equation (1), H ^T can represent the transpose matrix of the feature regression matrix. According to Equation (1), the image processing apparatus can calculate a position offset T _i (where i = 1, ..., N, N is a positive integer) corresponding to each of the N candidate selection regions sampled.

As another embodiment, there may be a case where a plurality of characteristic regression matrices H _j (where j = 1, ..., M, M is a positive integer) are learned in advance based on a plurality of sample frames. In this case, the image processing apparatus can repeatedly apply a plurality of feature regression matrices to each candidate selection region to calculate a plurality of position offsets. More specifically, the image processing apparatus can calculate the position offset T _i ^j corresponding to the candidate selection area X _i as shown in the following equation (2).

In Equation (2), q _i denotes a feature point existing in a candidate selection region X _i (where i = 1, ..., N, N is a positive integer), T _i ^j denotes an _i- ^th candidate selection region X _{The i-} th feature regression matrix is applied to _i to represent the calculated position offset. In Equation (2), H _j ^T denotes a transpose matrix of the j-th feature regression matrix. The image processing apparatus can calculate the _i- th position offset T _i corresponding to the i-th candidate selection area X _i based on the following equation (3).

The image processing apparatus can calculate an i-th position offset T _i corresponding to the i-th candidate selection region X _i by calculating an average value of the calculated position offsets using a plurality of feature regression matrices.

는

에 비례하도록 정의될 수 있다.In step 420, the image processing apparatus generates a plurality of position offsets T _i By applying a weight to each of them, a target position offset for tracking the target can be calculated. The image processing apparatus can determine the weight according to the variance V _i of the positional offset T _i ^j corresponding to the candidate selection area X _i . More specifically,

The

. ≪ / RTI >

Further, the image processing apparatus can determine that candidate selection regions existing at neighboring positions have similar weight. More specifically, the image processing apparatus can calculate the weighting matrix g that minimizes the target function f (g), and determine the weighting factor according to the weighting matrix g. The target function f (g) can be defined as Equation (4) below.

이고, D가 Diag{d₁, … , d_N}인 경우에, 상기 수학식 4의 라플라시안 행렬(Laplacian matrix) L은 D-Q로 계산될 수 있다. g=

는 각각의 후보 선택 영역의 위치에 따라 결정된 가중치의 가중 행렬을 나타내고,

을 나타낼 수 있다. 또한, 상기 수학식 4에서, 가중행렬 g는 각각의 원소가 0보다 작지 않도록 정의될 수 있다.m _ij may represent the overlapping rate of the i-th candidate selection area X _i and the j-th candidate selection area X _j . The overlap rate is the i-th candidate selection X _i and the j-th candidate selection, X _j is selected, the size and the i-th candidate for the region shared with the area X _i and the j-th candidate selection X _j size of the area with at least one account of the . Q can represent an N by N matrix with m _ij as an element. d _i is

, D is Diag {d ₁ , ... , d _N }, the Laplacian matrix L of Equation (4) can be calculated as DQ. g =

Represents a weighted matrix of weights determined according to the position of each candidate selection area,

Lt; / RTI > In Equation (4), the weighting matrix g may be defined such that each element is not smaller than zero.

Also, in step 420, the image processing apparatus can determine the positions of the candidate selection regions after the feature regression according to the positions of the respective candidate selection regions, the respective position offsets, and the weights. More specifically, the image processing apparatus can calculate the target position offset T for tracking the target as shown in Equation (5) below.

In Equation (5), X _i ^D represents a position vector representing a representative position of the i-th candidate selection region X _i . Accordingly, the position after the feature regression for the i-th candidate selection area X _i can be calculated as X _i ^D + T _i . The image processing apparatus can calculate the target position offset T after generalized regression using the weight g _i corresponding to each candidate selection region.

In step 430, the image processing apparatus may determine a final selection region containing the target using the target position offset T. [ The image processing apparatus can determine the position of each of the candidate selection regions after the regression based on the target position offset T. [ In one embodiment, the image processing apparatus can determine the last selected area in the current frame by applying the calculated target position offset T to the position of the previously stored base image.

Although not shown in Fig. 4, the image processing apparatus can store the second frame in which the final selection area is determined in the memory. In addition, when the number of frames stored in the memory exceeds a predetermined threshold value, the image processing apparatus can update the base image using the stored frame. More specifically, the image processing apparatus can select the base image to be newly updated based on the evaluation result value of the final selection area corresponding to each stored frame.

FIG. 5 is a flowchart illustrating a process of learning a characteristic regression matrix by an image processing apparatus according to another embodiment. Referring to FIG. 5, the image processing apparatus includes a step 510 of extracting a plurality of candidate selection regions using a base image of a sample frame, a step of calculating a position offset matrix between the extracted plurality of candidate selection regions and the base image Step 530 of learning a feature regression matrix on the input image based on step 520 and the calculated position offset matrix.

In step 510, the image processing apparatus can extract a plurality of candidate selection regions around the base image of the sample frame. In one embodiment, the image processing apparatus may sample N candidate selection regions of neighboring regions around the position of the base image. More specifically, the image processing apparatus can extract the plurality of candidate selection regions so that the position distribution of the candidate selection region follows an even distribution or a Gaussian distribution.

In another embodiment, the image processing apparatus can extract each candidate selection region having a fixed size. By way of example, one candidate selection region may be defined as a pixel block space defined by 32 by 32. Alternatively, the size of the candidate selection area may be the same as the size of the base image stored in advance.

과

로서 계산할 수 있다. 또한, 영상 처리 장치는 계산된 복수의 위치 오프셋을 이용하여 위치 오프셋 행렬 C=[C₁, C₂]를 계산해낼 수 있다. 보다 구체적으로,

으로 각각 정의될 수 있다.In step 520, the image processing apparatus can calculate the position offset matrix using the extracted positions of the candidate selection region and the base image. More specifically, the image processing apparatus can calculate the difference between the representative position of each of the plurality of candidate selection regions and the position of the base image as a position offset. Illustratively, the representative position of the candidate selection area may indicate the position of the center point in the candidate selection area. The image processing apparatus compares the representative position (x ₁ , y ₁ ) of the first candidate selection region with the position of the base image to calculate the first position offset (P ₁ ^x , P ₂ ^x ). Similarly, the image processing apparatus sets a plurality of position offsets of (x, y) coordinates corresponding to each of the plurality of candidate selection regions

and

. Further, the image processing apparatus can calculate the position offset matrix C = [C ₁ , C ₂ ] using a plurality of calculated position offsets. More specifically,

Respectively.

In operation 530, the image processing apparatus may learn a feature regression matrix for the input image based on the calculated position offset matrix. More specifically, the image processing apparatus can determine and store a feature regression matrix corresponding to a sample frame by applying the position offset matrix to feature points included in each candidate selection region and feature points in the base image.

The image processing apparatus can calculate the characteristic regression matrix H (h ₁ , h ₂ ) by minimizing the target function f (H) as shown in Equation (6) below.

In Equation (6), the feature point corresponding to the _i- th candidate selection region X _i is q _i , and the characteristic regression vectors for the X-coordinate and Y-coordinate are h ₁ and h ₂ , respectively. In one embodiment, the image processing apparatus may calculate the feature regression matrix H based on a logistic regression. The feature regression matrix H calculated by the image processing apparatus can be summarized as Equation (7) below.

In Equation (7), X represents a matrix relating to a representative position of the candidate selection region, and I represents a unit matrix having 1 on the diagonal and 0 as the remaining elements. The image processing apparatus can store the learned feature regression matrix in memory.

In another embodiment, the image processing apparatus may train and store M characteristic regression matrices using different M sample frames. In such a case, one sample frame corresponds to one feature regression matrix and M can represent a positive integer not less than two. The image processing apparatus according to the present embodiment can improve the accuracy of target tracking by more effectively removing outliers using a plurality of characteristic regression matrices. The trained and learned feature regression matrix can be used to track the target of any frame included in the input image. The detailed description of the target tracking process can be applied to the description of FIGS. 1 to 4 as it is, and a duplicate description will be omitted.

FIG. 6 is a flowchart illustrating a process of determining a final selection area including a target according to another embodiment of the present invention. Referring to FIG. 6, a method for determining a final selection region includes calculating (610) a similarity of each of a positive sample and a plurality of candidate selection regions with respect to a characteristic of a target, And determining (620) a final selection region that includes the target within the first region.

In operation 610, the image processing apparatus may calculate a degree of similarity that each of the plurality of candidate selection regions has with respect to the positive sample using a stored clustering model. Although not shown in FIG. 6, the image processing apparatus can determine an initial selection region for a target within an input frame, and extract a plurality of candidate selection regions around the determined initial selection region. The process of extracting a plurality of candidate selection regions by the image processing apparatus has been described in detail with reference to FIG. 3A and FIG. 3B, and therefore a duplicated description will be omitted.

More specifically, the clustering model may include a learned positive sample matrix by extracting features of a target region in a base image as positive samples. In addition, the clustering model may include a learned negative sample matrix by extracting a characteristic of a surrounding region existing within a predetermined range around the target region as negative samples. Illustratively, at least one of a gray image and a histogram of oriented gradients may be used as a feature of an image.

In one embodiment, in step 610, the image processing apparatus may use a sparse subspace clustering model (SSC), which is learned based on a plurality of base sub images, Can be calculated. More specifically, the sparse subspace clustering model may be implemented with a hybrid sparse subspace clustering model (HSSC) that includes a positive sample matrix and a negative sample matrix.

In operation 610, the image processing apparatus may calculate the similarity between the positive sample matrix included in the sparse partial spatial clustering model and the feature points included in each candidate selection region. In another embodiment, the image processing apparatus may calculate the similarity by comparing the positive sample matrix with feature points of a sub-region included in the candidate selection region. The image processing apparatus may add similarities with respect to a plurality of sub regions included in each of the candidate selection regions.

In addition, the image processing apparatus can determine the similarity degree corresponding to the candidate selection region based on the sum of similarities of the plurality of sub regions included in the candidate selection region. More specifically, the image processing apparatus can calculate an average value of similarities with respect to a plurality of sub-regions as the similarity of the candidate selection region.

In step 620, the image processing apparatus may determine the final selection area containing the target in the frame, according to the similarity degree corresponding to each candidate selection area. More specifically, the image processing apparatus can determine the candidate selection region having the maximum similarity to the positive sample among the plurality of candidate selection regions as the final selection region for the target. Further, the image processing apparatus may store the determined information of the final selection area as a target tracking result for the current frame. Illustratively, the information of the final selection region may include at least one of the size of the region, the position of the region, the image data in the region, and the feature data in the image.

Although not shown in FIG. 6, the image processing apparatus can compare the last selected area of the frame with previous frames stored in the memory. More specifically, the image processing apparatus can compare the similarity of the positive sample of the final selection region with the average of the similarity of the positive samples of the previous frames. As a result of comparison, when the similarity of the positive sample of the final selection area is larger than the similarity average value of the positive samples of the previous frames, the image processing apparatus can newly store the frame in which the final selection area is determined. In addition, the image processing apparatus can newly store the base image of the target within the frame having the maximum similarity to the positive sample among the previously stored frames.

FIGS. 7A and 7B are diagrams for explaining a process of extracting a positive sample and a negative sample using a sample frame according to another embodiment of the present invention. Referring to FIG. 7A, the image processing apparatus 700 receives a sample frame and outputs either a positive sample or a negative sample corresponding to the sample frame. The image processing apparatus 700 can be preliminarily learned to extract a positive sample or a negative sample within a specified sample frame. The image processing apparatus 700 can store each of the positive sample matrix 721 and the negative sample matrix 722 generated using the sample frame as a clustering model in the memory 710. [

Referring to FIG. 7B, the image processing apparatus 700 may extract feature points 731, 732, 733, 734, 735, 736, 737, and 738 in the input sample frame as positive samples. In one embodiment, the image processing apparatus 700 may extract, as the positive sample, feature points such as a face contour point of the target included in the base image of the sample frame, an eye point of the target, a mouse point of the target, In addition, the image processing apparatus 700 may sample a plurality of pixel points in a region 740 existing within a predetermined distance d based on feature points extracted from the positive sample, and extract the extracted pixel points as a negative sample. The predetermined distance d represents a variable designated so that the positive sample and the negative sample can be separated.

FIG. 8 is a flowchart illustrating a process of learning a sparse partial space clustering model by the image processing apparatus illustrated in FIG. 7A. Referring to FIG. 8, an image processing apparatus 700 includes a step 810 of calculating a coefficient matrix defining a sub-region of a positive sample using a positive sample matrix, and a spectral clustering using a calculated coefficient matrix Step 820 may be performed.

이고, 포지티브 샘플 행렬 A는

로 정의될 수 있다. 또한, 네거티브 샘플의 개수가 M 개인 경우, 네거티브 샘플은

로 정의될 수 있다. 영상 처리 장치(700)가 적어도 하나의 샘플 프레임으로부터 포지티브 샘플을 추출하고, 포지티브 샘플 행렬을 생성하는 과정에 대해서는 도 7a 및 도 7b와 함께 기재된 설명이 그대로 적용될 수 있어 중복되는 설명은 생략하기로 한다.In operation 810, the image processing apparatus 700 may optimize a production coefficient matrix using a predetermined positive sample matrix. If the number of positive samples is N,

, And the positive sample matrix A is

. ≪ / RTI > In addition, when the number of negative samples is M,

. ≪ / RTI > 7A and 7B can be directly applied to the process of extracting a positive sample from at least one sample frame and generating a positive sample matrix by the image processing apparatus 700, and a duplicated description will be omitted .

In operation 810, the image processing apparatus 700 may optimize the production coefficient matrix based on an LSR (Least Squares Regression) model. More specifically, the image processing apparatus 700 can calculate the optimal production coefficient matrix W ^* by minimizing the target function f (W) as shown in the following equation (8).

In Equation (8), W denotes a production coefficient matrix,? Denotes a constant, and | | _F represents an F matrix. The image processing apparatus 700 can calculate an optimal production coefficient matrix W ^* that minimizes f (W) defined according to Equation (8) as shown in Equation (9) below.

는

의 역행렬을 나타낸다. 영상 처리 장치(700)는 계산된 최적의 생산 계수 행렬 W^*을 아래의 수학식 10에 적용하여, 혼합계수행렬 B를 계산할 수 있다.In Equation (9), A ^T denotes a transposed matrix of the positive sample matrix,

The

≪ / RTI > The image processing apparatus 700 can calculate the mixing coefficient matrix B by applying the calculated optimal production coefficient matrix W ^* to the following equation (10).

In operation 820, the image processing apparatus 700 may perform spectral clustering using the calculated coefficient matrix. The image processing apparatus 700 can perform spectral clustering on the generated mixture coefficient matrix B and obtain a plurality of positive sample groups.

As another embodiment, the image processing apparatus 700 can perform spectral clustering on the mixture coefficient matrix B repeatedly a predetermined number of times, and obtain a plurality of positive sample groups. Illustratively, but not exclusively, the image processing apparatus 700 may perform spectral clustering until N positive samples are clustered into K positive sample groups, where K is an integer not greater than N ). The spectral clustering process is straightforward for technical experts and will not be described in detail.

The image processing apparatus 700 may count the number of times spectral clustering has been performed for the mixture coefficient matrix B as an index and determine whether to repeat the spectral clustering based on the counted index. More specifically, when the spectral clustering for the mixture coefficient matrix B is repeated K predetermined times, the image processing apparatus 700 stores the positive sample group generated in the K-th order in the sparse partial space clustering model, It is possible to terminate the iterative clustering.

As another embodiment, the image processing apparatus 700 may calculate the discriminantity coefficient matrix for determining whether the spectral clustering is repeated using the positive sample group and the negative sample. The image processing apparatus 700 can extract the positive sample and the negative sample corresponding to any kth positive sample group as one sample group. The image processing apparatus 700 can obtain the discriminating direction p _k corresponding to the positive sample in the sample group based on a predetermined graph embedding model. In the following description, the graph embedding model means a method of mapping one graph to another graph separated from the graph.

및 네거티브 샘플

이 존재하는 경우가 있을 수 있다. 이 경우에, 영상 처리 장치(700)는 두 샘플의 유클리드 거리를

로 계산하고, 상기 유클리드 거리

에 기반하여 가중치

를 계산할 수 있다. 일실시예로서 두 개의 샘플 모두가 포지티브 샘플이거나 네거티브 샘플인 경우, 영상 처리 장치(700)는 상기 두 개의 샘플 사이의 가중치를 0으로서 계산할 수 있다. 또한, 영상 처리 장치(700)는 계산된 가중치에 기초하여 그래프 임베딩에 이용되는 라플라시안 행렬을 계산하고, 상기 라플라시안 행렬에 기초하여 판별방향 p_k를 획득할 수 있다.More specifically, the image processing apparatus 700 can determine the weighting between the samples based on the Euclidean distance of the positive sample and the negative sample included in the sample group. Illustratively, the positive samples present in the kth group

And negative samples

There may be cases where the In this case, the image processing apparatus 700 calculates the Euclidean distance of two samples

, And the Euclidian distance

Based on the weight

Can be calculated. In one embodiment, if both samples are positive samples or negative samples, the image processing apparatus 700 may calculate the weight between the two samples as zero. In addition, the video processing device 700 may be based on the calculated weights to calculate the Laplacian matrix used in the graph embedding, and obtains the determined directions p _k on the basis of the Laplacian matrix.

과 포지티브 샘플 그룹의 평균값

의 유사성

을 계산할 수 있다.The image processing apparatus 700 can determine the similarity of the average value of the positive sample group and the positive sample group based on the discrimination direction p _k of each sample group. More specifically, the image processing apparatus 700 obtains a positive sample

And the average value of positive sample groups

Similarity of

Can be calculated.

에 기초하여 판별성 계수 행렬을 계산할 수 있다. 영상 처리 장치(700)는 포지티브 샘플

및 포지티브 샘플

사이의 판별성에 기초하여 아래의 수학식 12와 같이 유사도 계수

를 계산할 수 있다.In addition, the image processing apparatus 700 calculates the similarity

The discriminant coefficient matrix can be calculated. The image processing apparatus 700 includes:

And positive samples

On the basis of the discrimination between the similarity coefficients

Can be calculated.

와 제1 포지티브 샘플 그룹의 평균값

의 유사성을 나타내고, I_j ¹는 포지티브 샘플

와 제1 포지티브 샘플 그룹의 평균값

의 유사성을 나타낸다. 또한, 영상 처리 장치(700)는 상기 수학식 12에서 계산된 유사도 계수

를 원소로 하여 판별성 계수 행렬

를 획득할 수 있다.In Equation (12), I _i ¹ denotes a positive sample

And the average value of the first positive sample group

I _j ¹ represents the similarity of the positive sample

And the average value of the first positive sample group

Lt; / RTI > Further, the image processing apparatus 700 may calculate the similarity coefficient < RTI ID = 0.0 >

The discriminant coefficient matrix

Can be obtained.

Further, the image processing apparatus 700 can confirm the number of positive samples included in each positive sample group. If the number of identified positive samples is less than the predetermined threshold, the image processing apparatus 700 can determine the vacancy state of the positive sample group corresponding to the positive sample.

In addition, the image processing apparatus 700 may add a positive sample to the first positive sample group existing in a bare state. More specifically, the image processing apparatus 700 may supplement the positive sample of the second positive sample group having the degree of similarity equal to or greater than the predetermined threshold value to the first positive sample group with the first positive sample group. The image processing apparatus 700 may repeat the replenishment of the positive sample so that the number of positive samples in the first positive sample group is equal to or greater than the predetermined threshold value.

을 이용하여 아래의 수학식 13과 같이 혼합계수행렬 B를 획득할 수 있다.The image processing apparatus 700 includes an optimal production coefficient matrix W ^* and a discriminant coefficient matrix

The mixing coefficient matrix B can be obtained as shown in Equation (13) below.

는 상수를 나타낸다. 영상 처리 장치(700)는 혼합계수행렬 B에 대해 스펙트럴 클러스터링을 수행하고, 각각의 반복 차수에 대응하는 포지티브 샘플 그룹을 생성할 수 있다. 영상 처리 장치(700)는 상기 반복 차수가 미리 정의된 임계 차수에 도달할 때까지 상기 혼합계수행렬 B에 대한 스펙트럴 클러스터링을 수행할 수 있다.In Equation (13)

Represents a constant. The image processing apparatus 700 can perform spectral clustering on the mixture coefficient matrix B and generate a positive sample group corresponding to each repetition degree. The image processing apparatus 700 may perform spectral clustering on the mixture coefficient matrix B until the iteration order reaches a predefined critical order.

Also, the image processing apparatus 700 can perform Principal Component Analysis (PCA) on each of the generated positive sample groups and acquire sub regions of positive sample groups. Each sub-region of the positive sample group may be included in a sparse subspace clustering model. Also, each sub-region may include an average value of positive samples included in the sub-region.

The image processing apparatus 700 according to the present embodiment can generate a clustering model using positive samples with respect to the target. In addition, the image processing apparatus 700 may generate a sparse subspace clustering model corresponding to a sub-region in the positive sample, and may have robustness to image noise through the sparse subspace clustering model, Can be more accurately tracked. The sparse subspace clustering model may be learned from a plurality of sample frames based on the Euclidean distance of the positive sample with respect to the characteristic of the target and the negative sample with respect to the characteristic of the neighborhood of the target.

Although not shown in FIG. 8, the image processing apparatus 700 can update the sparse partial space clustering model. The image processing apparatus 700 can determine whether the last selection area is determined and the number of frames stored in the memory exceeds a predetermined threshold value. If the number of previously stored frames exceeds the threshold according to the determination result, the image processing apparatus 700 extracts a new sample frame from the input image and updates the sparse subspace clustering model using the extracted sample frame can do. More specifically, the image processing apparatus 700 can extract the sub-region of the positive sample group based on the newly extracted sample frame.

Also, the image processing apparatus 700 can update the base image based on the newly extracted sample frame. The image processing apparatus 700 can use the updated base image and the sparse subspace clustering model to track the target for a frame that is subsequently input. The method of FIG. 6 may be applied to the process of the image processing apparatus 700 to track the target, so that redundant description will be omitted.

9 is a block diagram illustrating an image processing apparatus according to another embodiment. 9, the image processing apparatus 900 may include an extracting unit 910, a first calculating unit 920, a second calculating unit 930, and a determining unit 940. The extracting unit 910 can extract a plurality of candidate selection regions from the input frame using the base image. The base image refers to an image of a region in which a target exists in a frame stored in the image processing apparatus 900. [ In one embodiment, the basis image may be included in a first frame of an input image including a target. In another embodiment, the base image may be included in an arbitrary frame of the input image as a final selection area in which the target is tracked by the image processing apparatus 900.

The first calculation unit 920 may track the target from feature points existing in each of the plurality of candidate selection regions. More specifically, the first calculation unit 920 may calculate a plurality of position offsets by applying a characteristic regression matrix to each of the plurality of candidate selection regions. Also, the first calculation unit 920 may calculate a target position offset by applying a weight to the calculated plurality of position offsets.

The second calculation unit 930 can calculate the similarity of each of the plurality of candidate selection regions and the positive samples related to the characteristics of the target. More specifically, the second calculation unit 930 may calculate the similarity according to a Hybrid Sparse Subspace Clustering (HSSC) model. More specifically, a mixed sparse subspace clustering model is learned using a positive sample for the target and a negative sample for the neighboring region of the target. The second calculation unit 930 can calculate the degree of similarity that the sub-region included in the candidate selection region has with respect to the positive sample, as shown in Equation (14) below.

는 각각의 부분공간에 포함되는 포지티브 샘플의 특징 평균값을 나타내고, r은 후보 선택 영역 내의 서브 영역의 인덱스를 나타낼 수 있다. 예시적으로, 상기 I는 상기 포지티브 샘플의 휘도값을 나타낼 수 있다.In Equation (14), U _k denotes the k-th subspace included in the mixed rare subspace clustering model, I denotes the characteristic of the positive sample included in each subspace,

Represents the feature average value of the positive samples included in each subspace, and r can represent the index of the sub region within the candidate selection region. Illustratively, I may represent the luminance value of the positive sample.

Also, the second calculation unit 930 may determine the similarity of the subspace having the maximum degree of similarity to the rth subspace among the subspaces included in the mixed sparse subspace clustering model as the similarity of the subspace r. More specifically, the second calculation unit 930 can calculate the similarity of the rth sub-region as shown in the following equation (15).

Also, the second calculation unit 930 may calculate the similarity corresponding to the _i- th candidate selection region X _i as shown in Equation (16) by adding the similarities of all the sub-regions included in the i-th candidate selection region.

The decision unit 940 may apply a first weight to the target position offset and apply a second weight to the similarity of each of the plurality of candidate selection regions to determine a final selection region containing the target. More specifically, the determination unit 940 can calculate evaluation information corresponding to the final selection area as shown in Equation (17) below.

는 후보 선택 영역의 최대 유사도이고,

은 타겟 위치 오프셋이고,

는 0보다 크고 1보다 작은 실수로서 가중 계수를 나타내고,

는 타겟을 포함하는 최종 선택 영역의 평가 정보이다.In Equation 17,

Is the maximum similarity of the candidate selection region,

Is the target position offset,

Represents a weighting factor as a real number greater than 0 and less than 1,

Is the evaluation information of the final selection area including the target.

The apparatus described above may be implemented as a hardware component, a software component, and / or a combination of hardware components and software components. For example, the apparatus and components described in the embodiments may be implemented within a computer system, such as, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable array (FPA) A programmable logic unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions. The processing device may execute one or more software applications that are executed on an operating system (OS) and an operating system. The processing device may also access, store, manipulate, process, and generate data in response to execution of the software. For ease of understanding, the processing apparatus may be described as being used singly, but those skilled in the art will recognize that the processing apparatus may have a plurality of processing elements and / As shown in FIG. For example, the processing unit may comprise a plurality of processors or one processor and one controller. Other processing configurations are also possible, such as a parallel processor.

The software may include a computer program, code, instructions, or a combination of one or more of the foregoing, and may be configured to configure the processing device to operate as desired or to process it collectively or collectively Device can be commanded. The software and / or data may be in the form of any type of machine, component, physical device, virtual equipment, computer storage media, or device , Or may be permanently or temporarily embodied in a transmitted signal wave. The software may be distributed over a networked computer system and stored or executed in a distributed manner. The software and data may be stored on one or more computer readable recording media.

The method according to an embodiment may be implemented in the form of a program command that can be executed through various computer means and recorded in a computer-readable medium. The computer readable medium may include program instructions, data files, data structures, and the like, alone or in combination. Program instructions to be recorded on the medium may be those specially designed and constructed for the embodiments or may be available to those skilled in the art of computer software. Examples of computer-readable media include magnetic media such as hard disks, floppy disks and magnetic tape; optical media such as CD-ROMs and DVDs; magnetic media such as floppy disks; Magneto-optical media, and hardware devices specifically configured to store and execute program instructions such as ROM, RAM, flash memory, and the like. Examples of program instructions include machine language code such as those produced by a compiler, as well as high-level language code that can be executed by a computer using an interpreter or the like. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

Although the embodiments have been described with reference to the drawings, various technical modifications and variations may be applied to those skilled in the art. For example, it is to be understood that the techniques described may be performed in a different order than the described methods, and / or that components of the described systems, structures, devices, circuits, Lt; / RTI > or equivalents, even if it is replaced or replaced. Therefore, other implementations, other embodiments, and equivalents to the claims are also within the scope of the following claims.

Claims (19)

Implemented by the processor:
A calculation unit for calculating a position offset with respect to each of the plurality of candidate selection areas in the second frame based on the position of the basis image in the first frame; And
Determining a final selection region including a target in the second frame according to the calculated position offset and a weight assigned to each of the plurality of candidate selection regions,
And the image processing apparatus.
The method according to claim 1,
Wherein the determination unit determines the weight based on a position in the second frame of each of the plurality of candidate selection areas.
The method according to claim 1,
Wherein the calculation unit calculates a plurality of position offsets by applying a characteristic regression matrix to each of the plurality of candidate selection regions and calculates a target position offset for calculating the target position offset by applying the weight to each of the plurality of position offsets, Processing device.
The method according to claim 1,
Wherein the calculation unit calculates a plurality of position offsets with respect to the first candidate selection area using a plurality of previously learned feature regression matrices and calculates a plurality of position offsets corresponding to the first candidate selection area using the average value of the plurality of position offsets An image processing apparatus for calculating a position offset.
5. The method of claim 4,
Wherein the plurality of feature regression matrices are learned based on a feature point in the base image and a position offset according to feature points of each of the plurality of sample frames.
The method according to claim 1,
An extraction unit for determining an initial selection area for the target in the second frame based on the base image in the first frame and extracting the plurality of candidate selection areas around the determined initial selection area,
Further comprising:
The method according to claim 6,
Wherein the extracting unit calculates an overall position offset between the first frame and the second frame and determines an initial selection area based on the calculated total position offset and position information in which the target exists in the base image, .
8. The method of claim 7,
Wherein the extraction unit extracts projection points corresponding to the feature points in the base image in the second frame and determines the overall position offset using the texture values of the determined projection points and points within a predetermined range Processing device.
9. The method of claim 8,
Wherein the extracting unit extracts points within a predetermined range around the determined projection point, determines a matching point corresponding to the feature point according to a similarity between a texture value of the extracted points and a texture value of the feature point, And the position of the matching point are compared with each other to determine the overall position offset.
The method according to claim 1,
A storage unit for storing the second frame in which the last selected area is determined and updating the base image according to a target tracking result value of the frames when the number of stored frames becomes equal to or greater than a threshold value,
Further comprising:
Calculating a similarity of each of a plurality of candidate selection regions and a positive sample with respect to a characteristic of the target; And
Determining a final selected region containing a target in a frame according to the calculated similarity
And an image processing method.
12. The method of claim 11,
Wherein the step of calculating the degree of similarity comprises:
Calculating the similarity by comparing features of the positive sample included in a Sparse Subspace Clustering (SSC) model with features of sub regions in each candidate selection region
And an image processing method.
13. The method of claim 12,
Wherein the step of calculating the degree of similarity comprises:
Calculating a degree of similarity of the first candidate selection region according to a sum of similarities with respect to a plurality of sub regions included in the first candidate selection region
And an image processing method.
13. The method of claim 12,
Wherein the sparse subspace clustering model is learned from a plurality of sample frames based on a Euclidean distance of a positive sample with respect to a characteristic of the target and a negative sample with respect to a characteristic of an adjacent region of the target.
12. The method of claim 11,
Comparing the similarity of the last selected region of the frame with the positive sample and the similarity average of the positive samples with previous frames; And
Storing the frame in accordance with the comparison result;
Further comprising the steps of:
16. The method of claim 15,
Comparing the number of stored frames to a predetermined threshold, and updating the sparse subspace clustering model using the stored frames as a sample frame according to a result of the comparison
Further comprising the steps of:
Computer-implemented,
An extracting unit for extracting a plurality of candidate selection regions from an input frame using a base image;
A first calculation unit for calculating a target position offset that tracks a target from feature points existing in each of the plurality of candidate selection areas;
A second calculation unit for calculating a similarity between each of the plurality of candidate selection regions and a positive sample related to a characteristic of the target; And
Determining a final selected region including the target by applying a first weight to the target position offset and applying a second weight to the similarity of each of the plurality of candidate selection regions,
And the image processing apparatus.
18. The method of claim 17,
Wherein the first calculator calculates a plurality of position offsets by applying a characteristic regression matrix to each of the plurality of candidate selection regions and calculates a target position offset by applying a weight to the calculated plurality of position offsets.
18. The method of claim 17,
The second calculator calculates the degree of similarity according to a Hybrid Sparse Subspace Clustering (HSSC) model learned using a positive sample of the target and a negative sample of a neighboring region of the target. Device.

Images