WO2012075729A1

WO2012075729A1 - Method and device for removing selected image in picture automatically

Info

Publication number: WO2012075729A1
Application number: PCT/CN2011/071073
Authority: WO
Inventors: 徐小明; 殷铭; 刘玉亭
Original assignee: 上海杰图软件技术有限公司
Priority date: 2010-12-08
Filing date: 2011-02-18
Publication date: 2012-06-14
Also published as: CN102005060B; CN102005060A

Abstract

A method and device for removing selected images in a picture automatically are provided, and the method includes: setting boundaries to be replaced in an original picture, so as to mask selected images; setting boundary-point-located blocks centered on the points on the boundaries to be replaced; finding a point p' with the highest block confidence among all of the boundary points, and the boundary-point-located block corresponding to the point p' being Ψ _p'; extending the boundaries to be replaced to form extended boundaries; finding a boundary-point-located block Ψ _q'having the largest feature similarity compared with the boundary-point-located block of the point p', the boundary point corresponding to Ψ _q' being q', and replacing Ψ _p' with Ψ _q'; assigning the maximum value of the feature similarity compared between eight neighbor pixels of the point q' and the point p' to the point p'; and returning to the step for searching the point with the highest confidence, and searching the point with the highest confidence in the unprocessed boundary points, until all of the boundaries to be replaced have been processed.

Description

Method and apparatus for automatically removing selected images in an image

The present invention relates to an image processing method and apparatus, and more particularly to a method and apparatus for automatically removing selected images appearing in an image. Background technique

In image shooting, it is often inevitable that partial images that are not desired to be saved appear in the image, which will cause great loss to the user's visual experience. The existing processing method is completed by manual removal. The specific method is to repair the local image area by means of an image editing tool and a manual image in the partial image to be removed. This type of operation is extremely labor intensive. Increased the burden of post processing.

Traditional repair image technologies fall into two categories. One type is the image inlaying technique based on geometric image model. This technique uses partial differential equations and physical thermal diffusion principles to repair images. Representatives are: M. Bertalmio^ AL Bertozzi et al. Navier-stokes, Fluid dynamics, and image and video inpainting; C. Ballester ^ V. Casel les et al "A variational model for fi ll ing - in gray level and color images"; TF Chan and J. Shen " Non- texture inpainting by curvature- Driven diffusions (CDD). This type of technique preserves the linear structure of the image, but is only suitable for patching small-scale defects in the image; it works very badly for slightly larger areas, often with significant blurring in the repair area. The other is image compretion technology based on texture synthesis. Among them are: L. Liang, C. Liu et al., Real-time texture synthesi s by patch-based sampling; A. Efros and WT Freeman "Image qui lting for texture synthes is and transfer"; M. Ashikhmin "Synthesizing natural textures"); This type of technical advantage is fast, but the shortcomings do not save the linear structure of the image well. Later authors have improved on the idea of inpainting, such as A. Criminisi et al. ((Object removal by exemplar-based inpainting); D. Simakov et al. (Summarizing visual data using bidirectional simi larity). CHO, TS (The post transform and its appl ications to image editing); KOMODAKIS, N. et al. ((Image completion using efficient bel ief propagation via priority scheduling ing and dynamic pruning); KUMAR, N., etc. « What is a good Nearest neighbors algorithm for finding simi lar patches in images?)) ; RUBINSTEIN, M. et al. "Improved seam carving for video retargeting"; WEXLER, Y and other "Space-time completion of video"). This technique has a good effect on filling large missing blocks in the image, and can also preserve the structure of the image, but the speed is very slow. For a picture of 1200*1200 size, it usually takes a few minutes or even half an hour to repair the 200*200 area. . Therefore, the above techniques have great limitations for local image removal or repair applications.

Therefore, how to automatically remove or repair selected image areas is an urgent problem to be solved. Summary of invention

SUMMARY OF THE INVENTION An object of the present invention is to solve the above problems and to provide a method for automatically removing selected images in an image, which has the advantages of fast repair speed and good repair effect.

Another object of the present invention is to provide an apparatus for automatically removing selected images from an image.

The technical solution of the present invention is: The present invention discloses a method for automatically removing selected images in an image, including:

Set the boundary to be replaced on the original image to cover the selected image;

Set the block where the boundary point centered on the point on the boundary to be replaced is located;

Find the point with the highest confidence of the block in all the boundary points / , the block where the point / corresponding boundary point is located

Ψ _Ρ ' ·,

Extending the boundary to be replaced to form an outer expanded boundary;

Find the block where the boundary point of the point / / is located in the range of the outer boundary of the outer expansion. The block where the feature similarity is the largest is ^, ^, and the corresponding boundary point is replaced by ^.

The maximum value of the feature similarity between the 8 adjacent pixels of the point q' and the point p' is assigned to the point p'; the step of finding the point with the highest confidence is found, and the confidence is found in the block where the unprocessed boundary point is located. The highest point until all the boundaries to be replaced have been processed.

An embodiment of the method for automatically removing selected images in an image according to the present invention further includes: before setting a boundary to be replaced on the original image:

The original image is reduced so that the subsequent processing object is the reduced original image.

In accordance with an embodiment of the method of automatically removing selected images from an image in accordance with the present invention, the original image is reduced by an interpolation algorithm.

An embodiment of a method of automatically removing selected images in an image in accordance with the present invention, finding points and points After that, according to the original image reduction ratio, find the corresponding two points p and ^ in the original image and the corresponding blocks and ^ of the two points p and ^, replace 3⁄4^ with ^, and 8 adjacent pixels and points of point g The feature similarity maximum value compared between ρ is assigned to point P.

According to an embodiment of the method for automatically removing selected images in an image according to the present invention, in the step of setting a boundary to be replaced on the original image, the boundary to be replaced is marked as a color different from the background color to make the background color The borders with different colors become the boundaries to be replaced.

The present invention also discloses an apparatus for automatically removing selected images in an image, including:

a boundary setting module, which sets a boundary to be replaced on the original image to cover the selected image;

a block setting module in which the boundary point is located, coupled to the boundary setting module, and set a block in which the boundary point centered on the point on the boundary to be replaced is located;

The confidence processing module is coupled to the confidence processing module to find the point with the highest confidence of the block in all the boundary points, and the block corresponding to the boundary point of the point // is ^

The boundary expansion module is coupled to the confidence processing module to expand the boundary to be replaced to form an outer boundary;

Alternatively the processing module, coupled to the outer boundary of the expansion module, the expansion of the outer boundary of the range to identify the boundary points and the point where the block // _Ψ Ρ 'feature similarity compared to the maximum point where the block boundary Ψ, ¥ _q' corresponding to The boundary point is q', and the block where the boundary point is located is replaced by the block 所在 where 边界_Ρ ';

The assignment module is coupled to the replacement processing module and the confidence processing module, and assigns the maximum value of the feature similarity between the 8 adjacent pixels of the point to the point/, and returns to the confidence processing module for processing, and is processed by the confidence degree. The module looks for the point with the highest confidence in the block where the unprocessed boundary points are located until all the boundaries to be replaced are processed.

An embodiment of the apparatus for automatically removing selected images in an image according to the present invention, the apparatus further comprising an image reduction module coupled to the boundary setting module, the image reduction module reducing the original image for subsequent processing The object is the original image after reduction.

In accordance with an embodiment of the apparatus for automatically removing selected images from an image in accordance with the present invention, an interpolation processing unit is provided in the image reduction module to reduce the original image by operation of an interpolation algorithm in the interpolation processing unit.

An embodiment of the apparatus for automatically removing selected images in an image according to the present invention further includes an original map mapping unit in the replacement processing module, the replacement processing module starting the original map mapping unit after finding the dot/and the dot ^ ', The original map mapping unit finds the corresponding two points ρ and g in the original image according to the original map reduction ratio and this Two points p and ^ correspond to the block and ^, so that ^ replaces 3⁄4^, and points in the assignment module

The maximum value of the feature similarity between the adjacent pixels and the point p is assigned to the point p.

According to an embodiment of the apparatus for automatically removing selected images in an image according to the present invention, in the boundary setting module, the boundary to be replaced is marked as a color different from the background color so that a boundary different from the background color is to be replaced. boundary. Compared with the prior art, the present invention has the following beneficial effects: the invention can automatically remove the image of the selected area in the image, so that the original image provider can more easily remove the image of the selected area, thereby solving the original image processing selected area. Image removal relies too much on manpower, inefficiency, and cost. DRAWINGS

1 is a flow chart of a first embodiment of a method of automatically removing selected images from an image of the present invention. 2 is a flow chart of a second embodiment of a method of automatically removing selected images from an image of the present invention. Figure 3 is a detailed flow chart of the embodiment of Figure 2.

4 is a schematic diagram of a first embodiment of an apparatus for automatically removing selected images from an image of the present invention. Figure 5 is a schematic diagram of a second embodiment of the apparatus for automatically removing selected images from an image of the present invention. Detailed description of the invention

The invention will now be further described with reference to the drawings and embodiments. First Embodiment of Method of Automatically Removing Selected Images in an Image FIG. 1 illustrates an embodiment of a method of automatically removing selected images in an image of the present invention. Referring to FIG. 1, the following is a detailed description of the steps of the method for automatically removing selected images in an image of the present embodiment.

Step S10: Set the boundary to be replaced on the original image to cover the selected image.

The border to be replaced can be marked as a uniform color, covering the background color on the original image. That is, the boundary to be replaced is marked as a color different from the background color, so that the boundary different from the background color can be judged as the boundary to be replaced.

Step S11: Set a block in which the boundary point centered on the point on the boundary to be replaced is located. In a certain pixel range around any point on the boundary to be replaced, a block centered on the point is set for subsequent comparison and replacement, and the size of the block can be arbitrarily set.

Step S12: Find the point with the highest confidence of the block among all the unprocessed boundary points, and the block where the corresponding boundary point is // is ί^.

Step S13: The boundary to be replaced is externally expanded to form an outer expanded boundary.

For example, if the original boundary to be replaced is (r X m), then the outer boundary is (r + areasize ) X (m ten areasize ), where areasize is a parameter indicating how many pixels are expanded outside the boundary to be replaced. Find block images that can be replaced in this range.

It should be noted that the unique pixel RGB value (that is, the color) set in step S10 for marking the boundary to be replaced is completely different from the RGB value of each pixel in the original image in the extended boundary range.

Step S14: Find the point with the highest degree of confidence in the range of the expanded boundary; / the block ^, compared with the block with the largest feature similarity, ^, ^, the corresponding boundary point is ^, with ^ To replace Ψ _Ρ ,

In this step, the image replacement of the block/location block is completed. However, in order to prevent pixel loss during replacement, a subsequent step is required to ensure that the replaced image is completely fused with the surrounding image.

Step S15: assigning the maximum value of the feature similarity between the adjacent pixels of the point ^ ' and the point // to the point

Ρ, °

After the end of this step, go to step Sl l until all the boundaries to be replaced are all queried.

In this step, it is also necessary to exclude the block in which the boundary point has been processed (for example, ψ _ρ ') from the search range of the next loop by a technical means, so as to ensure that it is not necessary to search again when jumping to step S11. Processed boundaries. Such a technical means is, for example, to set the confidence level of the point to 0 (see the embodiment of FIG. 3 for details).

In the present embodiment, a loop is performed between steps S11 to S15 until the boundary to be replaced described in step S10 is processed. Second Embodiment of Method of Automatically Removing Selected Images in an Image FIG. 2 illustrates a second embodiment of a method of automatically removing selected images in an image of the present invention. Referring to Figure 2, the following is a detailed description of the various steps in the method of the present embodiment. Step S20: The original image is reduced so that the subsequent processing object is the reduced original image. A size range (n X n) can be preset. If the original image itself is within this size range, no further scaling is required, and the original image is substituted for subsequent operations.

If the original image itself is not within this size range, an interpolation algorithm is employed, scaled to a size range of n X n , and the reduced image is stored separately. The purpose of downsizing is to increase the speed of the calculation, and to ensure that the number of traversals is greatly reduced when searching for related matching points and regions.

Step S21: Determine the boundary to be replaced after the original image is reduced.

Step S22: Set a block in which the boundary point centered on the point on the boundary to be replaced is located.

In a certain pixel range around any point on the boundary to be replaced, a block centered on the point is set for subsequent comparison and replacement, and the size of the block can be arbitrarily set.

Step S23: Find the point with the highest confidence of the block in all the unprocessed boundary points, and the block where the corresponding boundary point is // is ί^.

Step S24: The boundary to be replaced is externally expanded to form an outer expanded boundary.

It should be noted that the unique pixel RGB value (that is, the color) used to mark the boundary to be replaced in step S21 is different from the RGB value of each pixel in the original image in the extended boundary range.

Step S25: Find the point with the highest degree of confidence in the range of the expanded boundary; / the block ^, compared with the block where the feature similarity is the largest, ^, ^, the corresponding boundary point is ^, according to the original The graph is scaled down to find the corresponding two points p and g in the original graph and the blocks Ψρ and ¥ _q corresponding to the two points p and g, which are replaced by ^.

In this step, the image replacement of the block where the point p is located is completed. However, in order to prevent pixel loss during replacement, a subsequent step is required to ensure that the replaced image is completely fused with the surrounding image.

Step S26: Assign the maximum value of the feature similarity between the 8 adjacent pixels of the point q and the point p to the point p. After the end of this step, the process jumps to step S22 until all the boundaries to be replaced are all queried. In this step, there is also a need to use a technical means to block the block where the boundary point has been processed (for example

Ψ _Ρ ' ) is excluded from the search range of the next loop to ensure that the boundary that has been processed is not searched again when jumping to step S22. Such a technical means is, for example, to set the confidence level of the point to 0 (see the embodiment of FIG. 3 for details).

In the present embodiment, a loop is performed between steps S22 to S26 until the boundary to be replaced described in step S21 is processed. Refinement of the second embodiment of the method of automatically removing selected images in an image Fig. 3 shows a further refinement of the above-described embodiment of Fig. 2. Referring to Fig. 3, the following is a detailed description of each step in the refinement flow of the method of the second embodiment.

Step S30: Select the area to be repaired from the original image.

For example, if the original image is (W X H), the boundary to be repaired can be marked as a uniform color, covering the background color, and whether the boundary to be replaced is judged by the boundary on which the color is different from other colors.

Step S31: Determine whether both W and Η are greater than η.

η is a preset value, for example, 300 pixels, which is a suitable value to ensure the efficiency of the operation. If both W and Η are larger than η pixels, step S312 is performed, otherwise step S314 is performed.

Step S312: Acquire W and the maximum ratio of H to n is rate.

The formula used in this step is rate = max (W/n, H/n).

Step S314: Assign rate to 1.

Step S316: According to the ratio obtained in step S312, an interpolation algorithm is introduced into the original image, and the original image is reduced, and the reduction ratio is rate.

The interpolation algorithm used in this step can be bilinear, thereby reducing pixel loss.

Step S32: Perform a reduction operation using the formula ' =_^, / / ' = ^_ to obtain W' and H'.

Rate rate In this step, if the value is 1, according to step S314, the value is equal to Pf, and the value of // is equal to //. Step S33: Let block ^' size be t*t, ^' be a confidence level, and ^' value be 0.

In this step, the block ^ ' is a block with a side length of ?, which is a natural number, for example, the value is

10.

Step S34: Judging the point; ? ' is the area to be repaired as specified in step S30. The basis for determining in this step is whether the RGB value of the point is the pixel value of the area to be repaired. For example, if the pixel of the area to be repaired set in step S30 is green, then here is whether the RGB value of the point is green. If it is green, step S342 is performed, and if it is not green, step S344 is performed.

Step S342: Find the point;? The confidence level of the 'block ^'.

The calculation of the confidence of the block ^' is a well-known technique, such as:

P(p^ = C(p^D(p^))

a Here is the area of ^'. ^a is the image normalization factor, which defaults to 255. ^Ώ is the area to be repaired, w, is orthogonal to the point; 'Unit vector of tangent direction. For the ?? point of the illuminance line.

P

CO) represents the confidence value of point u, u represents a pixel which is determined ') of the point P' where the blocks _Ψ Ρ 'confidence.

Step S344: Find the next pixel in the reduced image (size '*H').

If there is a next pixel in this step, step S345 is performed, otherwise step S348 is performed.

Step S345: Set to the next pixel in the reduced image (size '*H'). After this step is completed, step S34 is performed.

Step S346: The confidence in step S342 confidence obtained in ηιαχψ _p Ψ _p and compared.

In this step, if the confidence level of Ψ _Ρ ' is greater than the confidence of max Ψ _ρ ', step S347 is performed, and if the confidence level of Ψ _ρ ' is less than or equal to the confidence of ηια χψ _ρ , step S35 is performed.

Step S348: The removal of the selected image is completed, and the method ends.

Step S347: Assign the confidence of ^' to the confidence of ax ί^', and then proceed to step S35. In this step, the save point is simultaneously used to bring the operation performed in step S37.

Step S35: If the pixel point is the last pixel found in the reduced image (size is '*H') If yes, go to step S36, otherwise go back to step S345 to find the next pixel.

Step S36: Extend the areos/ze pixels in the outermost rectangle of the current boundary curve, and define the outer region, and the size is (r+a easize)*(m+a easize).

In this step, the circumscribed rectangle is a rectangle * j formed by the outer edge of the area to be repaired as described in S30. In this step, the value of ar size is set to 35, for example, and this value can improve the efficiency of the subsequent steps when the operation is carried.

Step S37: Set a point in the area other than the area to be repaired in step S30 in the expanded area (size r+areo^eJYm+areo^e) in step S36. Let the block be the point-centered block area, the size of the block is t*t, and the sum of the squares of the Euclidean distance of ^ is infinite.

In this step, the value of the side length t of the block is set to, for example, 10, which is the same as the side length of the block ^'. It is convenient to replace the pixel of the block Ψρ' in the block q' in the subsequent step.

Step S38: It is judged whether there is still a block ^^ in the extended area, if yes, step S382 is performed, and if not, step S384 is performed. Step S382: judging the point and the point; the square distance sum of the 'European distance' (the magnitude of the sum of the squares of the Euclidean distance of l _q ^ _gb -I _p ^ _b f and min, if the square of the Euclidean distance of the point and the point is smaller than the European style If the distance is squared, then step S386 is performed, otherwise step S384 is performed, where /^ _3⁄4 represents the pixel value of q' point RGB, and _Ip , _rgb represents the pixel value of p' point RGB. Step S386: ji q'rgb

f is assigned to 欧式ηψ q, the sum of squared Euclidean distances. A point will also be retained in this step, which will be taken to the next step S387 or S388.

After this step is completed, the judgment of step S38 will be continued.

Step S384: If the rate value is equal to 1, proceed to step S387, otherwise proceed to step S388.

The rate value described in this step is the value obtained in step S312 or step S314.

Step S387: Copy the pixel of the block where the point is located to the point; ? 'in the block ^'.

This step will cause the pixels of the block where the point is to be replaced, and the removal of the selected image in this block is completed. After this step is completed, the subsequent step S392 will be performed.

Step S388: Find points and points in the original picture (size is V*ff). This step is a linear reflection The relationship is plotted as P = rate * /? ng = rate * '.

After this step is completed, step S389 will be performed.

Step S389: Copy the pixel of the block where the point q is located to the block Ψρ where the point P is located.

This step will replace the pixel of the block Ψ _Ρ in the block, and complete the removal of the selected image in this block. After this step is completed, step S394 will be performed.

Step S392: Assign the Euclidean squared minimum of the adjacent pixel of the point to the point of the ?? point. After this step is completed, the process proceeds to step S39.

Step S394: Assign the Euclidean squared minimum value of the 8 adjacent pixels of the point q to the P point to the point P. After this step is completed, the process proceeds to step S39.

Step S39: Set the confidence level of the point to 0.

This step makes the value judged in the subsequent step S346 always negative, so that the subsequent steps can be performed.

S35.

After this step is completed, step S344 will be performed. First Embodiment of Apparatus for Automatically Removing Selected Images in an Image FIG. 4 illustrates the principle of a first embodiment of an apparatus for automatically removing selected images from an image of the present invention. Referring to Figure 4, the following is a detailed description of the principles of the apparatus of the present embodiment.

The apparatus for automatically removing selected images in an image of the embodiment includes the following modules: a boundary setting module

10. The block setting module in the boundary point 1 1. The confidence processing module 12, the boundary expansion module 13, the replacement processing module 14, and the evaluation module 15. The connection relationship between the modules is: the boundary setting module 10 is coupled to the block setting module 1 where the boundary point is located, and the block setting module 1 1 where the boundary point is located is coupled to the confidence processing module 12, and the confidence processing module 12 is coupled to the boundary expansion. The module 13 , the boundary expansion module 13 is coupled to the replacement processing module 14 , and the replacement processing module 14 is coupled to the evaluation module 15 . The evaluation module is coupled to the confidence processing module 12 in addition to the replacement processing module 14 .

The boundary setting module 10 is configured to set a boundary to be replaced on the original image to cover the selected image. The border to be replaced can be marked as a uniform color, covering the background color on the original image. That is, the boundary to be replaced is marked as a color different from the background color, so that the boundary different from the background color can be determined as the edge to be replaced. Boundary.

The block setting module 11 in which the boundary point is located is used to set a block in which the boundary point centered on the point on the boundary to be replaced is located. In a certain pixel range around any point on the boundary to be replaced, a block centered on the point is set for subsequent comparison and replacement, and the size of the block can be arbitrarily set.

In the confidence processing module 12, the point with the highest confidence of the block is found among all the unprocessed boundary points; /, the block where the point/corresponding boundary point is located is ί^. The calculation of the block confidence is a conventional technical means and is described in detail in the refinement of the second embodiment of the method.

The boundary outer expansion module 13 expands the boundary to be replaced to form an outer expanded boundary. For example, if the original boundary to be replaced is (r X m), then the outer boundary is (r + areasize ) X (m+areasize ), where areasize is a parameter indicating how many pixels are expanded outside the boundary to be replaced. Find block images that can be replaced in this range.

It should be noted that the unique pixel RGB value (that is, the color) set in the boundary setting module 10 for marking the boundary to be replaced is completely different from the RGB value of each pixel in the original image in the extended boundary range.

In the replacement processing module 14, the block where the point with the highest degree of confidence is located in the range of the outer-expanded boundary is located, and the boundary point where the feature similarity is the largest is ^, ί^, and the corresponding boundary point is ^, ^, to replace i^.

In this module, the dot is completed; the image of the block is replaced, but in order to prevent pixel loss during replacement, subsequent module processing is required to ensure that the replaced image is completely fused with the surrounding image.

The assignment module 15 assigns the maximum value of the feature similarity between the 8 adjacent pixels of the point q' and the point p' to the point p'.

After processing by the evaluation module 15, the jump to the confidence processing module 12 is performed by the confidence processing module 12 to find the point with the highest confidence in the block of the unprocessed boundary point until all the boundaries to be replaced are all queried.

In the confidence processing module 12, it is necessary to exclude the block in which the boundary point has been processed (for example, ί^, ) from the search range of the next loop by a technical means to ensure that the boundary that has been processed is not found again. Such a technical means is, for example, to set the confidence level of the point to 0 (see the embodiment of Fig. 3 for details).

A second embodiment of an apparatus for automatically removing selected images in an image Figure 5 illustrates the principles of a second embodiment of the apparatus for automatically removing selected images from an image of the present invention. Referring to Figure 5, the following is a detailed description of the principles of the apparatus of the present embodiment.

The apparatus for automatically removing selected images in the image of the embodiment includes the following modules: an image reduction module 20, a boundary setting module 21, a block setting module 22 where a boundary point is located, a confidence processing module 23, a boundary expansion module 24, and a replacement. The processing module 25 and the evaluation module 26 are provided. The connection relationship between the modules is: the image reduction module 20 is coupled to the boundary setting module 21, the boundary setting module 21 is coupled to the block setting module 22 where the boundary point is located, and the block setting module 22 where the boundary point is located is coupled to the confidence processing module 23, The degree processing module 23 is coupled to the boundary expansion module 24, the boundary expansion module 24 is coupled to the replacement processing module 25, and the replacement processing module 25 is coupled to the evaluation module 26. The evaluation module 26 is coupled to the replacement processing module 25 to couple the confidence. Processing module 23. An interpolation processing unit 200 is provided in the image reduction module 20, and an original map mapping unit 250 is provided in the replacement processing module 25.

The image reduction module 20 reduces the original image so that the subsequent processing object is the reduced original image. In the image reduction module 20, a size range (n X n) can be preset. If the original image itself is within this size range, no further scaling is required, and the original image is substituted for subsequent operations.

If the original picture itself is not within this size range, the interpolation processing unit 200 is operated, and the interpolation processing unit 200 employs an interpolation algorithm to scale the original picture to the size range of n X n and separately store the reduced image. The purpose of downsizing is to increase the speed of the calculation, and to ensure that the number of traversals is greatly reduced when searching for related matching points and regions.

The boundary setting module 21 is configured to judge the boundary to be replaced after the original image is reduced to cover the selected image. The border to be replaced can be marked as a uniform color, covering the background color on the original image. That is, the boundary to be replaced is marked as a color different from the background color, so that the boundary different from the background color can be judged as the boundary to be replaced.

The block setting module 22 in which the boundary point is located is used to set a block in which the boundary point centered on the point on the boundary to be replaced is located. In a certain pixel range around any point on the boundary to be replaced, a block centered on the point is set for subsequent comparison and replacement, and the size of the block can be arbitrarily set.

In the confidence processing module 23, the point with the highest confidence of the block is found among all the unprocessed boundary points; /, the block where the point/corresponding boundary point is located is ί^. The calculation of the block confidence is a conventional technical means and is described in detail in the refinement of the second embodiment of the method.

The boundary outer expansion module 24 expands the boundary to be replaced to form an outer expanded boundary. For example, the original waiting The replacement boundary is (r X m), then the outer-expanded boundary is (r + areas i ze ) X (m+areas i ze ), where area i ze is a parameter indicating how many pixels are expanded outside the boundary to be replaced , will find the block image that can be replaced in this range.

It should be noted that the pixel-only RGB value (that is, the color) set in the boundary setting module 21 for marking the boundary to be replaced is completely different from the RGB value of each pixel in the original image in the extended boundary range.

In the replacement processing module 25, the block where the point with the highest degree of confidence is located in the range of the expanded boundary is located, and the block where the feature similarity is the largest, ^, ^, the corresponding boundary point is ^.

After the points and points are found, the original map mapping unit 250 finds the corresponding two points p and g in the original picture and the blocks and ^ corresponding to the two points according to the original picture reduction ratio. Replace with ^.

In this module, the image replacement of the block where the point ρ is located is completed. However, in order to prevent pixel loss during replacement, subsequent module processing is required to ensure that the replaced image is completely fused with the surrounding image.

The assignment module 26 assigns the maximum value of the feature similarity between the 8 adjacent pixels of the point q and the point p to the point

P °

After processing by the assignment module 26, a jump to the confidence processing module 23 is performed by the confidence processing module 23 to find the point with the highest confidence in the block of the unprocessed boundary point until all the boundaries to be replaced are all queried.

In the confidence processing module 23, it is necessary to use a technical means to block the block where the boundary point has been processed (for example, to exclude the search range of the next cycle, so as to ensure that the boundary that has been processed is not found again. Such technical means, for example. The block confidence of the point p is set to 0 (see the embodiment of Fig. 3 for details). The above embodiments are provided to those of ordinary skill in the art to implement or use the present invention, and those skilled in the art can In the case of the inventive concept of the present invention, various modifications and changes are made to the above-described embodiments, and thus the scope of the present invention is not limited by the above embodiments, but should be the largest in accordance with the innovative features mentioned in the claims. Scope.

Claims

Claim

1. A method for automatically removing selected images in an image, comprising:

Ψ _Ρ ' ·,

Extending the boundary to be replaced to form an outer expanded boundary;

2. The method for automatically removing selected images in an image according to claim 1, wherein before the boundary to be replaced on the original image is set, the method further comprises:

3. A method of automatically removing selected images from an image according to claim 2, wherein the original image is reduced by an interpolation algorithm.

4. The method for automatically removing selected images in an image according to claim 2, wherein after finding the point p' and the point q', finding corresponding two points in the original image according to the original image reduction ratio And ^ and the blocks and ^ corresponding to the two points P and g, replace the 3⁄4^ with ^, and assign the maximum value of the feature similarity between the 8 adjacent pixels of the point g and the point p to the point p.

5. The method of automatically removing selected images in an image according to claim 1, wherein in the step of setting a boundary to be replaced on the original image, marking the boundary to be replaced as a color different from the background color, The boundary that is different from the background color is made to be the boundary to be replaced.

6. A device for automatically removing selected images in an image, comprising:

The device for automatically removing selected images in an image according to claim 6, wherein the device further comprises an image reduction module, which is coupled to the boundary setting module, and the image reduction module reduces the original image. So that the subsequent processing object is the reduced original image.

8. The apparatus for automatically removing selected images in an image according to claim 7, wherein an interpolation processing unit is provided in the image reduction module, and the original image is reduced by operation of an interpolation algorithm in the interpolation processing unit. .

9. The apparatus for automatically removing selected images in an image according to claim 7, wherein the replacement processing module further includes an original map mapping unit that starts the original after finding the dot//and the point The map mapping unit, the original map mapping unit finds the corresponding two points ρ and g in the original image and the blocks and ^ corresponding to the two points ρ and g according to the original map reduction ratio, so that ^ is replaced by 3⁄4^, and is assigned In the module, the maximum value of the feature similarity between the 8 adjacent pixels of the point g and the point p is assigned to the point p.

10. The apparatus for automatically removing selected images in an image according to claim 6, wherein in the boundary setting module, the boundary to be replaced is marked as a color different from the background color so as to be different from the background color. The boundary becomes the boundary to be replaced.