CN103248795A - Method for estimating motion of image - Google Patents
Method for estimating motion of image Download PDFInfo
- Publication number
- CN103248795A CN103248795A CN2012100311815A CN201210031181A CN103248795A CN 103248795 A CN103248795 A CN 103248795A CN 2012100311815 A CN2012100311815 A CN 2012100311815A CN 201210031181 A CN201210031181 A CN 201210031181A CN 103248795 A CN103248795 A CN 103248795A
- Authority
- CN
- China
- Prior art keywords
- core block
- block
- image
- core
- dispersion
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a method for estimating motion of an image, which comprises the following steps: a plurality of core blocks in the present image are determined; on account of each core block, a search area of the core block is defined in a reference image and a first difference quantity between the core block and a first reference block and a second difference quantity between the core block and a second reference block are obtained; and according to the comparative result of both the first difference quantity and the second difference quantity of each core block, whether at least one reliable core block exists or not is judged and if at least one reliable core block exists, the mobile vector of the whole area of the current image is estimated according to the reliable core block.
Description
Technical field
The present invention relates to a kind of mobile evaluation method, relate in particular to and a kind ofly can effectively move the method for estimation at the image with repeat patterns.
Background technology
Along with the arriving of digital Age, the popularity rate of products such as digital camera or digital code camera also raises gradually.Compared to the image data that utilizes traditional camera or video camera to obtain shortcomings such as being difficult for improving preservation is arranged, utilize the obtained digitized image data of digital camera or video camera not only to preserve easily, also have and be convenient to follow-up editing and processing, and can advantage such as share by network.Manufacturer also focuses on function development more and progresses greatly in order to attract consumer's attention in the research and development of products such as digital camera, video camera, to satisfy consumer's various demand.For instance, the digital image at present obtains in the product, such as anti-hand shake, noise suppressed handles, or pan-shot etc. has belonged to very common function.
In above-mentioned functions, need mostly to utilize mobile estimation to produce preferable shooting results.For example, anti-hand shakes and noise suppressed handle to need reliably that mobile estimation result carry out motion compensation, and image to move the accuracy of estimation will be successfully to shoot the key of stable full-view image.Generally speaking, the mobile estimation of image is to utilize the resulting All Ranges motion-vector of mobile estimating is carried out in the zone to estimate the universe motion-vector.Yet this kind mode obtains insecure universe motion-vector estimation result easily.When particularly many repeat patterns (repeated pattern) occurring in image, more difficult generation is the universe motion-vector accurately.
Summary of the invention
In view of this, the invention provides a kind of image and move evaluation method, can reduce the pattern that repeats in the image and estimate the negative effect that causes for moving.
The present invention proposes a kind of image and moves evaluation method, and the method comprises a plurality of cores (kernel) block that determines in the present image.At each core block, the Search Area of this core block of definition in the reference image, and obtain this core block respectively and first reference block in the Search Area and first measures of dispersion between second reference block and second measures of dispersion.According to the comparative result of first measures of dispersion and second measures of dispersion of each core block, judge in all core blocks, whether there is at least one reliable core block.If there is at least one reliable core block, then estimate the universe motion-vector of present image according to above-mentioned reliable core block.
In one embodiment of this invention, wherein determine the step of the above-mentioned core block in the present image to comprise according to a plurality of default fixed position in the image at present and decide the core block.
In one embodiment of this invention, wherein determine the step of the core block in the present image to comprise that present image is carried out characteristic value obtains program to determine the core block.
In one embodiment of this invention, wherein at each core block, the Search Area of definition core block in the reference image, and obtain the core block respectively and the step of first reference block in the Search Area and first measures of dispersion between second reference block and second measures of dispersion comprise that the absolute difference that calculates position corresponding pixel value in core block and first reference block sums up (Sum of Absolute Difference, SAD) with as first measures of dispersion, wherein the position of first reference block in the reference image and the core block position in image at present is identical.And the absolute difference that calculates position corresponding pixel value in core block and second reference block is summed up with as second measures of dispersion, and wherein second reference block is block the most similar to the core block in Search Area.
In one embodiment of this invention, wherein according to the comparative result of first measures of dispersion and second measures of dispersion of each core block, whether judgement exists the step of reliable core block to comprise at each core block in all core blocks, calculates first measures of dispersion of this core block and ratio and the difference of second measures of dispersion.If ratio more than or equal to second preset value, judges then that corresponding core block belongs to reliable core block more than or equal to first preset value and difference.
In one embodiment of this invention, this image moves evaluation method and also comprises at each core block, if the position of core block in image at present is identical therewith the most similar position of block in the reference image, core block therewith in Search Area, judge that then this core block belongs to reliable core block.
In one embodiment of this invention, wherein estimate that according to reliable core block the step of the universe motion-vector of present image comprises that whether the quantity of judging reliable core block is more than or equal to the 3rd preset value.If then utilize above-mentioned reliable core block to divide other regional motion-vector to estimate the universe motion-vector of present image.
In one embodiment of this invention, this image move that evaluation method also comprises if the quantity of reliable core block less than the 3rd preset value, or in all core blocks, do not have reliable core block, then with the universe motion-vector of default motion-vector as present image.
In one embodiment of this invention, wherein default motion-vector is null vector.
Based on above-mentioned, the present invention system according to each block in the image with reference to the measures of dispersion between the identical and the most similar block in position in the image, determine when estimating the universe motion-vector of this image, whether will include the regional motion-vector of this block in consideration according to this.Just can have the accuracy that promotes mobile estimation under the situation of repeat patterns at image thus.
For above-mentioned feature and advantage of the present invention can be become apparent, embodiment cited below particularly, and conjunction with figs. is described in detail below.
Description of drawings
Fig. 1 is the flow chart that moves evaluation method according to the shown image of one embodiment of the invention;
Fig. 2 is the schematic diagram according to a plurality of core blocks of the shown present image of one embodiment of the invention;
Fig. 3 A is according to the shown present image of one embodiment of the invention and the schematic diagram of one of them core block;
Fig. 3 B is according to the shown reference image of one embodiment of the invention and the schematic diagram of Search Area;
Fig. 4 is the flow chart that moves evaluation method according to the shown image of another embodiment of the present invention.
Reference numeral:
S110～S150: the described image of one embodiment of the invention moves each step of evaluation method;
K1～K20, K: core block;
300a: present image;
C: location of pixels;
300b: with reference to image;
SA: Search Area;
RB1: first reference block;
RB2: second reference block;
M: the length and width of Search Area;
S410～S460: the described image of another embodiment of the present invention moves each step of evaluation method.
Embodiment
Fig. 1 is the flow chart that moves evaluation method according to the shown image of one embodiment of the invention.The image of present embodiment moves evaluation method and is applicable to as various video capturing devices such as digital camera or digital code cameras, in order to move estimation at the captured image of video capturing device, the motion-vector information that estimates then is applicable to the multiple function that video capturing device provides, shake function, noise suppressed processing capacity of anti-hand for example, or pan-shot (sweep panorama) function etc., and can promote the effect of above-mentioned functions.
See also Fig. 1, at first shown in step S110, in the captured present image of video capturing device, determine a plurality of cores (kernel) block.Present embodiment is to decide the core block according to a plurality of default fixed position in the present image.For instance, as shown in Figure 2, present embodiment is to define 20 core blocks (that is core block K1 to K20) according to 20 fixed positions that default in the image.That is no matter video capturing device is which kind of scene shot to go out present image at, and the number of core block and the position in present image all can keep fixing, promotes treatment effeciency thereby this kind mode can reduce computational complexity.Yet the present invention is not limited thereto, in another embodiment, also can obtain program and decide the core block by present image being carried out characteristic value.Wherein, it can be that bent angle detects (corner detection) program etc. that characteristic value is obtained program, is not limited at this.Obtaining position, core block that program determines and quantity by characteristic value can be different along with the included scenery of present image, therefore can reflect image feature.
Then in step S120, at each core block, one with reference to image in the definition this core block Search Area, and obtain first measures of dispersion between first reference block in this core block and the Search Area, and obtain second measures of dispersion between second reference block in this core block and the Search Area.In detail, the core block is big or small identical with first and second reference block, and the position of first reference block in the reference image is identical with the position of core block in present image, and second reference block then is block the most similar to the core block in Search Area.
Please consult Fig. 3 A and Fig. 3 B simultaneously, the size of Search Area is preset value (for example, the length and width of Search Area is m, are the block of pixels of big or small m * m) in the present embodiment.Be example with the core block K among the present image 300a, location of pixels C is the central point of core block K in present image 300a, is the corresponding Search Area SA of core block K and put obtained m * m block of pixels centered by location of pixels C in reference image 300b.Because core block K's is big or small known, as long as therefore obtain the position (for example, obtain the location of pixels C of central point) of core block K in present image 300a, just can obtain the first reference block RB1 in the same position of reference image 300b.In addition, can utilize as universe search (full search), diamond search (diamond-based search), or hexagonal search various search algorithm methods such as (hexagon-based search) are found out in Search Area SA with the most alike block of core block K and are used as the second reference block RB2.What must specify is that the present invention is not limited the kind of search algorithm method.In the present embodiment, first measures of dispersion is absolute difference sum total (the Sum of Absolute Difference of position corresponding pixel value among core block K and the first reference block RB1, SAD), second measures of dispersion then is the absolute difference sum total of position corresponding pixel value among core block K and the second reference block RB2.The size of supposing each core block among the present image 300a is n * n pixel, and present embodiment for example is to calculate the first measures of dispersion Diff1 and the second measures of dispersion Diff2 with following formula:
Wherein, Kp represents that pixel value, RB1p and the RB2p of pixel P in the K of core block then represent the pixel value of pixel P in the first reference block RB1 and the second reference block RB2 respectively.Because the second reference block RB2 is block the most alike with core block K in Search Area SA, the second measures of dispersion Diff2 is the minimum difference value of calculating among the Search Area SA, and the distance of the first reference block RB1 and the second reference block RB2 is the regional motion-vector of core block K.Yet mandatory declaration is, also can adopt in other embodiments other modes calculate core block K respectively and the first reference block RB1 and the second reference block RB2 between measures of dispersion.
After calculating corresponding first measures of dispersion in each core block and second measures of dispersion, next shown in step S130, according to the comparative result of first measures of dispersion and second measures of dispersion of each core block, judge in all core blocks, whether there is at least one reliable core block.Specifically, present embodiment is to go to calculate first measures of dispersion of core block and ratio and the difference of second measures of dispersion at each core block, if ratio more than or equal to second preset value, judges then that corresponding core block belongs to reliable core block more than or equal to first preset value and difference.Wherein, first preset value can have different set points with the kind of video capturing device or the function of subsequent applications with second preset value.In another embodiment, if the most similar position of block in the reference image, core block is identical therewith in the position of a core block in present image and the Search Area, then this core block also can directly be judged to be reliable core block.
If there is not any reliable core block in all core blocks, then shown in step S140, the image of present embodiment moves evaluation method can be assumed to the motion-vector of present image one default motion-vector.For instance, default motion-vector for example is null vector, and it represents that present image is not mobile.
If yet in all core blocks, there is more than one reliable core block, shown in step S150, estimate the universe motion-vector of present image according to all reliable core blocks.Present embodiment is the universe motion-vector that the regional motion-vector that utilizes each reliable core block (that is the measures of dispersion between the most similar block in reliable core block and the Search Area) is estimated present image.Specifically, from the regional motion-vector of all reliable core blocks, produce the universe motion-vector of present image by a filter.For example, calculate the mean value of regional motion-vector of all reliable core blocks or the universe motion-vector that median is used as present image.
As shown in Figure 1, present embodiment can't adopt the regional motion-vector of all core blocks to estimate the universe motion-vector of present image.The substitute is and calculate earlier each core block and with reference to each other measures of dispersion between two reference block in the image, above-mentioned two measuress of dispersion of recycling are inapplicable regional motion-vector rejecting.Thus, even if many repeat patterns occurring in the image at present, move evaluation method by image shown in Figure 1 and also can obtain comparatively accurate and reliable universe motion-vector.
Fig. 4 is the flow chart that moves evaluation method according to the shown image of another embodiment of the present invention.In the present embodiment, the quantity of reliable core block can influence the estimation result of the universe motion-vector of present image.Yet, because the step S410 of Fig. 4 is same or similar to step S130 to the step S110 of step S430 and Fig. 1, so do not repeat them here.
See also the step S430 of Fig. 4, if the judged result of step S430 is for being (namely, in all core blocks, have more than one reliable core block), then shown in step S450, judge that whether the quantity of reliable core block is more than or equal to the 3rd preset value.In the present embodiment, the big I of the 3rd preset value adopts different set points according to the subsequent applications of the universe motion-vector of present image.
If the quantity of reliable core block is less than the 3rd preset value, then shown in step S440, with the universe motion-vector of default motion-vector as present image.For instance, default motion-vector can be null vector.
Yet, if the quantity of reliable core block shown in step S460, utilizes all reliable core blocks to divide other regional motion-vector to estimate the universe motion-vector of present image more than or equal to the 3rd preset value.Similarly, present embodiment is the universe motion-vector that produces present image by a filter from the regional motion-vector of all reliable core blocks.
By each step shown in Figure 4, then can when the quantity of reliable core block is very few, avoid because cause judging by accident the situation of universe motion-vector with the regional motion-vector estimation of negligible amounts.
In sum, image shown in the present moves evaluation method when the universe motion-vector of the present image of estimation, can inapplicable regional motion-vector be rejected according to the core block and with reference to the measures of dispersion between the reference block in the image, increase the accuracy of estimation universe motion-vector accordingly.Particularly have under the situation of many repeat patterns at present image, move evaluation method by image of the present invention and also can produce the result of universe motion-vector estimation reliably.In addition, image of the present invention moves evaluation method can decide the core block with default fixed position, thereby can save operation time.Further, the core block of use required for the present invention and the measures of dispersion between the reference block are the calculation procedures that just need carry out when determining regional motion-vector originally, therefore need not expend extra operation time, promote treatment effeciency and can reduce computational complexity.
Though the present invention discloses as above with embodiment, so it is not in order to limiting the present invention, and any person of an ordinary skill in the technical field when can doing a little change and retouching, and does not break away from the spirit and scope of the present invention.
Claims (9)
1. an image moves evaluation method, and this method comprises:
Determine a plurality of core blocks in the present image;
At each those core block, one with reference to image in a Search Area of this core block of definition, and one first measures of dispersion and one second measures of dispersion between one first reference block that obtains this core block respectively and in this Search Area and one second reference block;
According to this first measures of dispersion of each those core block and a comparative result of this second measures of dispersion, judge in those core blocks, whether there is at least one reliable core block; And
If then a universe motion-vector of this present image is estimated in this at least one reliable core block of basis.
2. image according to claim 1 moves evaluation method, wherein determines the step of those core blocks in this present image to comprise:
Decide those core blocks according to a plurality of default fixed position in this present image.
3. image according to claim 1 moves evaluation method, wherein determines the step of those core blocks in this present image to comprise:
This present image is carried out a characteristic value obtain program to determine those core blocks.
4. image according to claim 1 moves evaluation method, wherein at each those core block, this with reference to image in this Search Area of this core block of definition, and this first measures of dispersion between this first reference block that obtains this core block respectively and in this Search Area and this second reference block and the step of this second measures of dispersion comprise:
The absolute difference sum total of calculating position corresponding pixel value in this core block and this first reference block is with as this first measures of dispersion, and wherein this first reference block is identical with the position of this core block in this present image with reference to the position in the image at this; And
This absolute difference sum total of calculating position corresponding pixel value in this core block and this second reference block is with as this second measures of dispersion, and wherein this second reference block is block the most similar to this core block in this Search Area.
5. image according to claim 1 moves evaluation method, wherein according to this first measures of dispersion of each those core block and this comparative result of this second measures of dispersion, judge in those core blocks, whether to exist the step of this at least one reliable core block to comprise:
At each those core block, calculate this first measures of dispersion of this core block and a ratio and a difference of this second measures of dispersion; And
If this ratio more than or equal to one second preset value, judges then that this core block belongs to reliable core block more than or equal to one first preset value and this difference.
6. image according to claim 1 moves evaluation method, wherein also comprises:
At each those core block, if in this Search Area the block the most similar to this core block at this with reference to the position in the image, identical with the position of this core block in this present image, judge that then this core block belongs to reliable core block.
7. image according to claim 1 moves evaluation method, wherein estimates that according to this at least one reliable core block the step of this universe motion-vector of this present image comprises:
Judge that whether the quantity of this at least one reliable core block is more than or equal to one the 3rd preset value; And
If then utilize this at least one reliable core block to divide other regional motion-vector to estimate this universe motion-vector of this present image.
8. image according to claim 7 moves evaluation method, wherein also comprises
If there is not reliable core block in the quantity of this at least one reliable core block less than the 3rd preset value or in those core blocks, then with a default motion-vector this universe motion-vector as this present image.
9. image according to claim 8 moves evaluation method, should default motion-vector be null vector wherein.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210031181.5A CN103248795B (en) | 2012-02-13 | 2012-02-13 | Image moves evaluation method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210031181.5A CN103248795B (en) | 2012-02-13 | 2012-02-13 | Image moves evaluation method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103248795A true CN103248795A (en) | 2013-08-14 |
| CN103248795B CN103248795B (en) | 2016-03-30 |
Family
ID=48928002
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210031181.5A Active CN103248795B (en) | 2012-02-13 | 2012-02-13 | Image moves evaluation method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103248795B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101065964A (en) * | 2004-09-27 | 2007-10-31 | 德克萨斯仪器股份有限公司 | Motion stabilization |
| US20080240241A1 (en) * | 2007-03-27 | 2008-10-02 | Nao Mishima | Frame interpolation apparatus and method |
-
2012
- 2012-02-13 CN CN201210031181.5A patent/CN103248795B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101065964A (en) * | 2004-09-27 | 2007-10-31 | 德克萨斯仪器股份有限公司 | Motion stabilization |
| US20080240241A1 (en) * | 2007-03-27 | 2008-10-02 | Nao Mishima | Frame interpolation apparatus and method |
Non-Patent Citations (1)
| Title |
|---|
| M.H.SHAKOOR等: "Fast Digital Image Stabilization By Motion Vector Prediction", 《2010 2ND CONFERENCE ON ENVIRONMENTAL SCIENCE AND INFORMATION APPLICATION TECHNOLOGY》, 31 December 2010 (2010-12-31), pages 151 - 154 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103248795B (en) | 2016-03-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11538232B2 (en) | Tracker assisted image capture | |
| KR102150776B1 (en) | Face location tracking method, apparatus and electronic device | |
| US9852511B2 (en) | Systems and methods for tracking and detecting a target object | |
| US10311579B2 (en) | Apparatus and method for detecting foreground in image | |
| US8130277B2 (en) | Method and system for intelligent and efficient camera motion estimation for video stabilization | |
| CN101080930B (en) | Motion vector calculation device and motion vector calculation method | |
| JP2021534600A (en) | Video stabilization methods and equipment, as well as non-temporary computer-readable media | |
| CN104144345B (en) | Carry out real time imaging at mobile terminal and know method for distinguishing and this mobile terminal | |
| CN107483953B (en) | Inter frame motion estimation method, apparatus and electronic equipment | |
| CN103744974B (en) | Method and device for selecting local interest points | |
| CN106504265B (en) | Estimation optimization method, equipment and system | |
| Hung et al. | A fast algorithm of temporal median filter for background subtraction. | |
| CN107767399B (en) | Forklift monitoring method and monitoring system | |
| CN104754180A (en) | Method and device for reducing video frame jitter | |
| CN104717402A (en) | Space-time domain joint noise estimation system | |
| KR101806453B1 (en) | Moving object detecting apparatus for unmanned aerial vehicle collision avoidance and method thereof | |
| CN103248795B (en) | Image moves evaluation method | |
| JPWO2011162309A1 (en) | Object region extraction apparatus, method, and program | |
| JP4289170B2 (en) | Noise amount measuring apparatus and video receiver | |
| CN111684489B (en) | Image processing method and device | |
| TWI502979B (en) | Method of image motion estimation | |
| CN105959531A (en) | Moving image detection system and method | |
| CN102215321B (en) | Mobile detection method and device | |
| EP3798976B1 (en) | Method and system for determining dynamism in a scene by processing a depth image | |
| US8111292B2 (en) | Method and apparatus for stably correcting handshake in camera panning situation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20161220 Address after: China Taiwan Hsinchu Science Park Hsinchu City Road No. 12 5 floor Patentee after: Glomerocryst semiconductor limited company Address before: Chinese Taiwan Hsinchu City Science Park force Road No. 12 Patentee before: Huajing Science & Technology Co., Ltd. |
|
| TR01 | Transfer of patent right |