CN102054159A - Method and device for tracking human faces - Google Patents
Method and device for tracking human faces Download PDFInfo
- Publication number
- CN102054159A CN102054159A CN2009102085107A CN200910208510A CN102054159A CN 102054159 A CN102054159 A CN 102054159A CN 2009102085107 A CN2009102085107 A CN 2009102085107A CN 200910208510 A CN200910208510 A CN 200910208510A CN 102054159 A CN102054159 A CN 102054159A
- Authority
- CN
- China
- Prior art keywords
- image
- face
- center
- people
- color
- 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
- Studio Devices (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method for intelligently tracking human faces, which relates to the computer technical field. The method comprises the following steps of: (1) obtaining an initialized image of a human face to be tracked; (2) gathering statistics of pixel distribution characteristics of the initialized image, wherein the pixel distribution characteristics at least comprise a color histogram of the initialized image and an edge histogram of the initialized image; (3) obtaining a color probability image of the human face to be traced according to the color histogram; (4) adjusting the center of the color probability image to the gravity center of the color probability image; (5) obtaining a human face tracking window according to the edge histogram of the initialized image and an edge histogram of the adjusted color probability image; and (6) taking the acquired human face tracking window as an initialized image of the human face to be tracked of the next frame of image. The invention correspondingly provides a device for intelligently tracking human faces. By the adoption of the technical proposal disclosed in the invention, more accurate video tracking images with human faces as the main part can be obtained for IM videos.
Description
Technical field
The present invention relates to field of computer technology, particularly a kind of method and apparatus of face tracking.
Background technology
IM (Instant Messaging, instant messaging) develops into today, is accepted by most netizen user.No matter the user is in life or work, all can use IM software to realize and friend that colleague and classmate, client's etc. exchanges and communication make IM software become one of indispensable instrument in user's daily life gradually in a large number.
In numerous purposes of IM software, the IM video is used by most netizen because of its advantage such as convenient, fast.Existing IM video can be understood as: A and B carry out Video chat, and A is the person of being taken, and B is the observer.Wherein, B can observe the image of A by network, and the observed image of B is the image that the video capture equipment of A is directly gathered, the not image through adjusting.There is following shortcoming in this existing IM video: in some cases, when the party A-subscriber from video capture equipment situations such as main scope far away or that do not taking, the effect of the image of the viewed A of user B just can not be fine, and for example people's face is fuzzy, distance does not far see Chu etc.For these situations, it is the display mode of video main contents that user B can select with people's face, when the IM of user A client receives the above-mentioned selection signal of user B, automatically start the face tracking technology, network communication between A and the B need not to transmit the original size frame, but can transmit the facial image of custom size (usually less than original size), therefore can reduce the transmission data.Like this, even also can reach smooth relatively video effect in the occasion of network environment difference; And can the dynamic head portrait of people's face dynamic image of acquisition as user A in the IM client window of user B will be followed the tracks of simultaneously.
The main stream approach of existing face tracking technical field has particle filter (Partic1e Filtering) and two kinds of technology of Mean Shift; Particle filter is bigger owing to influenced by number of particles, makes that speed is low relatively when being applied to the face tracking of video window, and therefore when long-distance video was chatted, picture was discontinuous, was difficult to use; Iteration speed is very fast for Mean Shift algorithm (is example with Intel OpenCV storehouse), but it is based on features of skin colors, makes that the approaching neck neck zone of the colour of skin is often detected easily mistakenly, therefore can not obtain human face region more accurately, as shown in Figure 1.
Summary of the invention
For in the IM video, obtain more accurately, based on the video image of people's face, the embodiment of the invention provides a kind of method and apparatus of face tracking.Described technical scheme is as follows:
A kind of method of face tracking, described method comprises:
Obtain the initialisation image of people's face to be tracked;
Add up the pixel distribution feature of described initialisation image, described pixel distribution feature comprises the color histogram of described initialisation image and the edge histogram of described initialisation image at least;
Obtain the color probabilistic image of described people's face to be tracked according to described color histogram;
The center of described color probabilistic image is adjusted to the center of gravity of described color probabilistic image, obtains adjusted color probabilistic image;
According to the edge histogram of described initialisation image and the edge histogram of described adjusted color probabilistic image, obtain the window of face tracking;
With the window of the face tracking of described acquisition initialisation image as people's face to be tracked of next frame image.
The initialisation image of described acquisition people's face to be tracked specifically comprises:
Adopt Detection for Moving Target,, obtain approximate head zone according to human head and shoulder portion proportionate relationship;
Adopt people's face detection sorter that the image of described approximate head zone is carried out intelligent people's face detection, obtain the initialisation image of people's face to be tracked.
After the approximate head zone of described acquisition, described method also comprises:
Described approximate head zone is amplified;
Accordingly, described employing people face detection sorter carries out intelligent people's face detection to the image of described approximate head zone, specifically comprises:
Adopt people's face to detect sorter the approximate head zone after amplifying is carried out intelligent people's face detection.
Described center with described color probabilistic image is adjusted to the center of gravity of described color probabilistic image, specifically comprises:
Calculate the center of gravity of described color probabilistic image;
Judge that whether difference between described center and the described center of gravity is smaller or equal to predefined threshold values;
If the center adjusted of then described color probabilistic image is to the center of gravity of described color probabilistic image;
If not, described center is moved a step-length according to predefined step-length towards described center of gravity, and carry out judge described center and described center of gravity difference whether smaller or equal to the step of predefined threshold values.
According to the edge histogram of described initialisation image and the edge histogram of adjusted described color probabilistic image, obtain the window of face tracking, specifically comprise:
Add up the edge histogram of adjusted described color probabilistic image, and the edge histogram of the described initialisation image of normalization;
Utilize the edge histogram of the later initialisation image of described normalization, the edge histogram calculating Pasteur distance with adjusted color probabilistic image obtains the highest people's face window of similarity degree.
Described similarity degree is the highest to be specially described Pasteur's distance smaller or equal to preset value.
A kind of device of face tracking, described device comprises:
Initialisation image obtains module, is used to obtain the initialisation image of people's face to be tracked;
Pixel distribution characteristic statistics module is used to add up the pixel distribution feature of described initialisation image, and described pixel distribution feature comprises the color histogram of described initialisation image and the edge histogram of described initialisation image at least;
The color probabilistic image obtains module, is used for obtaining according to described color histogram the color probabilistic image of described people's face to be tracked;
The center adjusting module is used for the center of described color probabilistic image is adjusted to the center of gravity of described color probabilistic image, obtains adjusted color probabilistic image;
The face tracking window obtains module, is used for according to the edge histogram of described initialisation image and the edge histogram of described adjusted color probabilistic image, obtains the window of face tracking; With the window of the face tracking of described acquisition initialisation image as people's face to be tracked of next frame image.
Described initialisation image obtains module and specifically comprises:
Approximate head zone obtains the unit, is used to adopt Detection for Moving Target, according to human head and shoulder portion proportionate relationship, obtains approximate head zone;
Initialisation image obtains the unit, is used to adopt people's face detection sorter that the image of described approximate head zone is carried out intelligent people's face detection, obtains the initialisation image of people's face to be tracked.
Initialisation image obtains module and also comprises:
Amplifying unit is used for the approximate head zone that described approximate head zone acquisition unit obtains is amplified;
Accordingly, described initialisation image obtains the unit and specifically is used for: adopt people's face to detect sorter and the approximate head zone after amplifying is carried out intelligent people's face detect.
The center adjusting module specifically comprises:
Computing unit is used to calculate the center of gravity of described color probabilistic image;
Judging unit is used to judge that whether difference between described center and the described center of gravity is smaller or equal to predefined threshold values;
The center adjustment unit is not if the judged result that is used for described judging unit when being, is adjusted described center; If the judged result of described judging unit for not the time, moves a step-length according to predefined step-length towards described center of gravity with described center, and starts judging unit.
The face tracking window obtains module and specifically comprises:
Computing unit is used to add up the edge histogram of adjusted described color probabilistic image, and the edge histogram of the described initialisation image of normalization;
Obtain the unit, be used to utilize the edge histogram of the later initialisation image of normalization, the edge histogram calculating Pasteur distance with adjusted color probabilistic image obtains the highest people's face window of similarity degree.
The beneficial effect that the technical scheme that the embodiment of the invention provides is brought is:
Obtain initial human face region by intelligent human face detection tech, and by statistics human face region pixel distribution feature, it is the color statistical information, and change the difference of iterative computation itself and regional center according to regional barycenter, zone after obtaining changing, and, ask for people's face outline map for the people's face figure that obtains in real time, carry out the masterplate coupling by the zone after every frame and the variation, obtain the final accurate human face region of every frame, and real-time update people face masterplate, finally obtained comparatively accurate face tracking result, overcome the shortcoming of prior art.
Description of drawings
Fig. 1 uses the result schematic diagram that prior art is carried out face tracking;
Fig. 2 is the method flow diagram of the face tracking that provides in the embodiment of the invention 1;
Fig. 3 is the method flow diagram of the face tracking that provides in the embodiment of the invention 2;
Fig. 4 uses the result schematic diagram that prior art is carried out face tracking in the embodiment of the invention 2;
Fig. 5 utilizes the method for the face tracking that provides in the embodiment of the invention 2 to carry out the result schematic diagram of face tracking;
Fig. 6 is the device synoptic diagram of the face tracking that provides in the embodiment of the invention 3.
Embodiment
For making the purpose, technical solutions and advantages of the present invention clearer, embodiment of the present invention is described further in detail below in conjunction with accompanying drawing.
Embodiment 1
For in the IM video, obtain more accurately, based on the video image of people's face, present embodiment provides a kind of method of face tracking, referring to Fig. 2, this method comprises:
201: the initialisation image that obtains people's face to be tracked;
Wherein, can adopt Detection for Moving Target,, obtain approximate head zone according to human head and shoulder portion proportionate relationship; The image that adopts people's face to detect sorter pairing approximation head zone again carries out intelligent people's face and detects, and obtains the initialisation image of people's face to be tracked.
After obtaining approximate head zone, this method also comprises: will be similar to head zone and amplify;
Accordingly, adopt people's face to detect sorter the more approximate head zone after amplifying is carried out intelligent people's face detection.
202: the pixel distribution feature of statistics initialisation image, the pixel distribution feature comprises the color histogram of initialisation image and the edge histogram of initialisation image at least;
203: the color probabilistic image that obtains people's face to be tracked according to color histogram;
204: the center of color probabilistic image is adjusted to the center of gravity of color probabilistic image, obtains adjusted color probabilistic image;
Comprising:
Calculate the center of gravity of color probabilistic image;
Judge that whether difference between center and the center of gravity is smaller or equal to predefined threshold values;
If then the center adjusted of color probabilistic image is to the center of gravity of color probabilistic image;
If not, the center is moved a step-length according to predefined step-length towards center of gravity, and carry out judge center and center of gravity between difference whether smaller or equal to the step of predefined threshold values.
205:, obtain the window of face tracking according to the edge histogram of initialisation image and the edge histogram of adjusted color probabilistic image; With the window of the face tracking that obtains initialisation image as people's face to be tracked of next frame image.
Add up the edge histogram of adjusted color probabilistic image, and the edge histogram of normalization initialisation image;
Utilize the edge histogram of the later initialisation image of normalization, the edge histogram calculating Pasteur distance with adjusted color probabilistic image obtains the highest people's face window of similarity degree.
Wherein, the highest Pasteur's distance that is specially of similarity degree is less than preset value.
Obtain initial human face region by intelligent human face detection tech, and by statistics human face region pixel distribution feature, it is the color statistical information, and change the difference of iterative computation itself and regional center according to regional barycenter, zone after obtaining changing, and, ask for people's face outline map for the people's face figure that obtains in real time, carry out the masterplate coupling by the zone after every frame and the variation, obtain the final accurate human face region of every frame, and real-time update people face masterplate, finally obtained comparatively accurate face tracking result, overcome the shortcoming of prior art.
Embodiment 2
For in the IM video, obtain more accurately, based on the video image of people's face, the embodiment of the invention provides a kind of method of face tracking, is that the main frame at IM client place is an example with the executive agent, referring to Fig. 3, this method comprises:
301: the video image that collects is in real time carried out intelligent people's face detect, obtain initial human face region;
Wherein, carrying out the real-time equipment of gathering of video image can be video camera, the first-class equipment with shoot function of shooting.
Need to prove that intelligent people's face detects, and is in current input image or video and detects human face region.Comparatively ripe at present human face detection tech mainly adopts people's face Haar feature description to realize, the fast face detection algorithm based on face characteristic that calendar year 2001 Viola proposes is widely adopted, and is integrated into Intel OpenCV storehouse (the BSD agreement of increasing income).The synoptic diagram that people's face detects can be as shown in Figure 4.In embodiments of the present invention, detect sorter with the people's face that adopts Intel OpenCV to realize and realize the detection of intelligent people's face, thereby obtain initial human face region.
Also need to prove, the video image that capture apparatus collected for low resolution, the embodiment of the invention is before carrying out intelligent people's face detection to it, adopted Detection for Moving Target, the utilization background subtracts method and detects moving region (human body) in the camera form, according to the proportionate relationship of human head and shoulder portion, obtain approximate head zone; And after amplifying for the approximate head zone that obtains, adopt above-mentioned sorter to carry out intelligent people's face again and detect.The beneficial effect of doing like this is:
(1) not high or face characteristic is not under the outstanding especially situation, almost can't detect people's face to the people's face sorter among the former OpenCV in resolution; And the means such as moving object detection by in advance can progressively be dwindled sensing range, and with multiple dimensioned processing thought, be detected on than large scale, can obtain current human face region immediately;
(2) the people's face sorter among the former OpenCV detects based on face distribution characteristicss such as people's face portion symmetries, and when certain background (in the experiment as the cupboard of tape square, mask etc.) and this feature of people's face relatively were similar in the camera scene, detection was easy to make a mistake; And the algorithm of utilization motion detection has directly been got rid of these static target contexts, therefore can detect true man more accurately.
302: the pixel distribution feature of adding up initial human face region;
Wherein, the pixel distribution feature of adding up initial human face region specifically can comprise: calculate the color histogram of initial human face region, and add up its edge histogram, color histogram is changed into the color probabilistic image.
Wherein, need to prove, can make the calculating of carrying out color histogram with the following method, for example: original color image is gone to HSV (hue, saturation, value, tone, saturation degree, lightness) color space, extract the tone component, add up this histogram of component.Can adopt following method to carry out the statistics of edge histogram, for example: initial facial image is gone to the single channel image, calculate edge image, add up this image histogram.
Further, color histogram being converted into the color probabilistic image specifically refers to:
With RGB (red, green, blue, red, green, blue) image transitions of color space arrives the hsv color space, and extracts the tone component of image in the hsv color space, according to the tone component statistic histogram that extracts, then histogram is carried out back projection, promptly obtain the color probabilistic image.Each pixel value is the aggregate-value of tone component correspondence in histogram among the former HSV in the color probabilistic image.Therefore, in initial human face region, more if certain color value occurs, then its reaction in probabilistic image is comparatively obvious.
303: the centre of gravity place of color probabilistic image in the calculating pixel distribution characteristics;
Wherein, need to prove that the center of gravity of color probabilistic image obtains according to the moment characteristics of image-region, reacted this color of image aggregation extent.Because color probabilistic image centre of gravity place can use a lot of prior aries to calculate, and does not give unnecessary details so do not do.
304: the center of color probabilistic image is adjusted to its centre of gravity place;
Wherein, need to prove that the center of image specifically refers to the central point of image frame; The center of color probabilistic image is adjusted to its centre of gravity place is specifically as follows following operation:
Judge that whether distance (being difference) between center and the centre of gravity place is smaller or equal to predefined threshold values;
If, but illustrate that the center of color probabilistic image is adjusted to centre of gravity place in the permissible error scope, then continue execution in step 305;
If not, but the center that the color probabilistic image is described is not adjusted to centre of gravity place as yet in the permissible error scope, then the center is moved towards centre of gravity place with predefined step-length, whenever move a unit step-length, return and carry out the difference judged between center and the centre of gravity place whether smaller or equal to the step of predefined threshold values.
For example in fact, the center position coordinates of the color probabilistic image of filling in colors on a sketch is for (x1, y1), centre of gravity place is that (x2, y2), then predefined step-length is that (dx, dy), wherein, dx, dy both can be positive numbers, can be again negatives.At first calculate (x1, y1) with (x2, y2) whether the difference between smaller or equal to predefined threshold values, if then execution in step 305; If not, then with the center towards (x2 y2) moves a unit step-length, this moment the coordinate of center promptly change to (x1+dx, y1+dy), calculate (x1+dx, y1+dy) with (x2, y2) whether the difference between smaller or equal to predefined threshold values; If smaller or equal to, then execution in step 305; If greater than, then continue the center towards (x2 y2) moves a unit step-length, and so repeatedly, the difference between center and center is smaller or equal to predefined threshold values.
305: statistics is adjusted the edge histogram of the color probabilistic image behind the center, and the people's face edge histogram that obtains in the normalization step 302;
Wherein, need to prove that normalization people face edge histogram specifically refers to: the number of times that the number of times that each pixel value is occurred occurs divided by all pixels and, obtain the normalization edge histogram.
306: carry out the iteration coupling in the new edge histogram that the people's face edge histogram that obtains after the normalization is obtained in step 305, obtain the highest facial image of similarity degree;
Wherein, can use the similarity degree between histogram Pasteur distance calculation normalized people's face edge histogram and the new edge histogram, when Pasteur distance during smaller or equal to a certain preset value, normalized people's face edge histogram is with newly the similarity degree between the edge histogram is the highest.
307: the facial image window that the similarity degree of acquisition is the highest continues to use the intelligent-tracking that this method is carried out people's face in the video image as the home window of next frame image.
Fig. 5 is for using the tracking results that intelligent face tracking method obtained that the embodiment of the invention provides, and as can be seen from Figure 5, by scheme that should the embodiment of the invention provided, can be more accurately the video scene of chat camera be carried out face tracking.
The embodiment of the invention provides a kind of face tracking method, this method obtains initial human face region by intelligent human face detection tech, and by statistics human face region pixel distribution feature, it is the color statistical information, and change the difference of iterative computation itself and regional center according to regional barycenter, zone after obtaining changing, and, ask for people's face outline map for the facial image that obtains in real time, carry out the masterplate coupling by the zone after every frame and the variation, obtain the final accurate human face region of every frame, and real-time update people face masterplate, finally obtain comparatively accurate face tracking result, overcome the shortcoming of prior art; And, when under complex environment, adopting intelligent human face detection tech, also adopted Detection for Moving Target, further improved accuracy rate and speed that people's face detects.
Embodiment 3
For in the IM video, obtain more accurately, based on the video image of people's face, the embodiment of the invention provides a kind of device of face tracking, referring to Fig. 6, this device comprises:
Initialisation image obtains module 601, is used to obtain the initialisation image of people's face to be tracked;
Pixel distribution characteristic statistics module 602 is used to add up the pixel distribution feature of initialisation image, and the pixel distribution feature comprises the color histogram of initialisation image and the edge histogram of initialisation image at least;
The color probabilistic image obtains module 603, is used for obtaining according to color histogram the color probabilistic image of people's face to be tracked;
The face tracking window obtains module 605, is used for according to the edge histogram of initialisation image and the edge histogram of adjusted color probabilistic image, obtains the window of face tracking; With the window of the face tracking that obtains initialisation image as people's face to be tracked of next frame image.
Wherein, under a kind of embodiment, initialisation image obtains module 601 and specifically comprises:
Approximate head zone obtains the unit, is used to adopt Detection for Moving Target, according to human head and shoulder portion proportionate relationship, obtains approximate head zone;
Initialisation image obtains the unit, and the image that is used to adopt people's face to detect sorter pairing approximation head zone carries out intelligent people's face detection, obtains the initialisation image of people's face to be tracked.
Further, initialisation image acquisition module 601 also comprises:
Amplifying unit is used for the approximate head zone that approximate head zone acquisition unit obtains is amplified;
Accordingly, initialisation image acquisition unit specifically is used for: adopt people's face to detect sorter the approximate head zone after amplifying is carried out intelligent people's face detection.
Wherein, under a kind of embodiment, center adjusting module 604 specifically comprises:
Computing unit is used to calculate the center of gravity of color probabilistic image;
Judging unit is used for judging that whether difference between center and the center of gravity is smaller or equal to predefined threshold values;
The center adjustment unit is not if the judged result that is used for judging unit when being, is adjusted the center; If the judged result of judging unit for not the time, moves a step-length according to predefined step-length towards center of gravity with the center, and starts judging unit.
Wherein, under a kind of embodiment, the face tracking window obtains module 605 and specifically comprises:
Computing unit is used to add up the edge histogram of adjusted color probabilistic image, and the edge histogram of normalization initialisation image;
Obtain the unit, be used to utilize the edge histogram of the later initialisation image of normalization, the edge histogram calculating Pasteur distance with adjusted color probabilistic image obtains the highest people's face window of similarity degree.
Wherein, the highest Pasteur's distance that refers to of similarity degree is smaller or equal to preset value.
The above only is preferred embodiment of the present invention, and is in order to restriction the present invention, within the spirit and principles in the present invention not all, any modification of being done, is equal to replacement, improvement etc., all should be included within protection scope of the present invention.
Claims (11)
1. the method for a face tracking is characterized in that, described method comprises:
Obtain the initialisation image of people's face to be tracked;
Add up the pixel distribution feature of described initialisation image, described pixel distribution feature comprises the color histogram of described initialisation image and the edge histogram of described initialisation image at least;
Obtain the color probabilistic image of described people's face to be tracked according to described color histogram;
The center of described color probabilistic image is adjusted to the center of gravity of described color probabilistic image, obtains adjusted color probabilistic image;
According to the edge histogram of described initialisation image and the edge histogram of described adjusted color probabilistic image, obtain the window of face tracking;
With the window of the face tracking of described acquisition initialisation image as people's face to be tracked of next frame image.
2. the method for face tracking as claimed in claim 1 is characterized in that, the initialisation image of described acquisition people's face to be tracked specifically comprises:
Adopt Detection for Moving Target,, obtain approximate head zone according to human head and shoulder portion proportionate relationship;
Adopt people's face detection sorter that the image of described approximate head zone is carried out intelligent people's face detection, obtain the initialisation image of people's face to be tracked.
3. the method for face tracking as claimed in claim 2 is characterized in that, after the approximate head zone of described acquisition, described method also comprises:
Described approximate head zone is amplified;
Accordingly, described employing people face detection sorter carries out intelligent people's face detection to the image of described approximate head zone, specifically comprises:
Adopt people's face to detect sorter the approximate head zone after amplifying is carried out intelligent people's face detection.
4. the method for face tracking as claimed in claim 1 is characterized in that, described center with described color probabilistic image is adjusted to the center of gravity of described color probabilistic image, specifically comprises:
Calculate the center of gravity of described color probabilistic image;
Judge that whether difference between described center and the described center of gravity is smaller or equal to predefined threshold values;
If the center adjusted of then described color probabilistic image is to the center of gravity of described color probabilistic image;
If not, described center is moved a step-length according to predefined step-length towards described center of gravity, and carry out judge described center and described center of gravity difference whether smaller or equal to the step of predefined threshold values.
5. the method for face tracking as claimed in claim 1 is characterized in that, and is described according to the edge histogram of described initialisation image and the edge histogram of adjusted described color probabilistic image, obtains the window of face tracking, specifically comprises:
Add up the edge histogram of adjusted described color probabilistic image, and the edge histogram of the described initialisation image of normalization;
Utilize the edge histogram of the later initialisation image of normalization, the edge histogram calculating Pasteur distance with adjusted color probabilistic image obtains the highest people's face window of similarity degree.
6. the method for face tracking as claimed in claim 5 is characterized in that, described similarity degree is the highest to be specially described Pasteur's distance smaller or equal to preset value.
7. the device of a face tracking is characterized in that, described device comprises:
Initialisation image obtains module, is used to obtain the initialisation image of people's face to be tracked;
Pixel distribution characteristic statistics module is used to add up the pixel distribution feature of described initialisation image, and described pixel distribution feature comprises the color histogram of described initialisation image and the edge histogram of described initialisation image at least;
The color probabilistic image obtains module, is used for obtaining according to described color histogram the color probabilistic image of described people's face to be tracked;
The center adjusting module is used for the center of described color probabilistic image is adjusted to the center of gravity of described color probabilistic image, obtains adjusted color probabilistic image;
The face tracking window obtains module, is used for according to the edge histogram of described initialisation image and the edge histogram of described adjusted color probabilistic image, obtains the window of face tracking; With the window of the face tracking of described acquisition initialisation image as people's face to be tracked of next frame image.
8. the device of face tracking as claimed in claim 7 is characterized in that, described initialisation image obtains module and specifically comprises:
Approximate head zone obtains the unit, is used to adopt Detection for Moving Target, according to human head and shoulder portion proportionate relationship, obtains approximate head zone;
Initialisation image obtains the unit, is used to adopt people's face detection sorter that the image of described approximate head zone is carried out intelligent people's face detection, obtains the initialisation image of people's face to be tracked.
9. the device of face tracking as claimed in claim 8 is characterized in that, initialisation image obtains module and also comprises:
Amplifying unit is used for the approximate head zone that described approximate head zone acquisition unit obtains is amplified;
Accordingly, described initialisation image obtains the unit and specifically is used for: adopt people's face to detect sorter and the approximate head zone after amplifying is carried out intelligent people's face detect.
10. the device of face tracking as claimed in claim 7 is characterized in that, the center adjusting module specifically comprises:
Computing unit is used to calculate the center of gravity of described color probabilistic image;
Judging unit is used to judge that whether difference between described center and the described center of gravity is smaller or equal to predefined threshold values;
The center adjustment unit is not if the judged result that is used for described judging unit when being, is adjusted described center; If the judged result of described judging unit for not the time, moves a step-length according to predefined step-length towards described center of gravity with described center, and starts judging unit.
11. the device of face tracking as claimed in claim 7 is characterized in that, the face tracking window obtains module and specifically comprises:
Computing unit is used to add up the edge histogram of adjusted described color probabilistic image, and the edge histogram of the described initialisation image of normalization;
Obtain the unit, be used to utilize the edge histogram of the later initialisation image of normalization, the edge histogram calculating Pasteur distance with adjusted color probabilistic image obtains the highest people's face window of similarity degree.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200910208510.7A CN102054159B (en) | 2009-10-28 | 2009-10-28 | Method and device for tracking human faces |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200910208510.7A CN102054159B (en) | 2009-10-28 | 2009-10-28 | Method and device for tracking human faces |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102054159A true CN102054159A (en) | 2011-05-11 |
| CN102054159B CN102054159B (en) | 2014-05-28 |
Family
ID=43958461
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN200910208510.7A Active CN102054159B (en) | 2009-10-28 | 2009-10-28 | Method and device for tracking human faces |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102054159B (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102833486A (en) * | 2012-07-05 | 2012-12-19 | 深圳泰山在线科技有限公司 | Method and device for real-time adjusting face display scale in video image |
| CN103795851A (en) * | 2012-10-31 | 2014-05-14 | Lg电子株式会社 | Mobile terminal and method for controlling the same |
| WO2016090522A1 (en) * | 2014-12-12 | 2016-06-16 | Xiaoou Tang | Method and apparatus for predicting face attributes |
| CN105760826A (en) * | 2016-02-03 | 2016-07-13 | 歌尔声学股份有限公司 | Face tracking method and device and intelligent terminal. |
| CN105913028A (en) * | 2016-04-13 | 2016-08-31 | 华南师范大学 | Face tracking method and face tracking device based on face++ platform |
| CN106339072A (en) * | 2015-07-09 | 2017-01-18 | 中国科学院声学研究所 | Distributed large data real-time processing system and method based on left and right brain model |
| CN106709932A (en) * | 2015-11-12 | 2017-05-24 | 阿里巴巴集团控股有限公司 | Face position tracking method and device and electronic equipment |
| CN108256532A (en) * | 2018-01-10 | 2018-07-06 | 武汉斗鱼网络科技有限公司 | Image processing method, device, electronic equipment and storage medium |
| CN113673461A (en) * | 2021-08-26 | 2021-11-19 | 深圳随锐云网科技有限公司 | Method and device for realizing selection of human face and human figure region based on 4K + AI |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1492379A (en) * | 2002-10-22 | 2004-04-28 | 中国科学院计算技术研究所 | Method for covering face of news interviewee using quick face detection |
| CN101377813A (en) * | 2008-09-24 | 2009-03-04 | 上海大学 | Method for real time tracking individual human face in complicated scene |
-
2009
- 2009-10-28 CN CN200910208510.7A patent/CN102054159B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1492379A (en) * | 2002-10-22 | 2004-04-28 | 中国科学院计算技术研究所 | Method for covering face of news interviewee using quick face detection |
| CN101377813A (en) * | 2008-09-24 | 2009-03-04 | 上海大学 | Method for real time tracking individual human face in complicated scene |
Non-Patent Citations (3)
| Title |
|---|
| 张宏志等: "基于CamShift的目标跟踪算法", 《计算机工程与设计》 * |
| 王亮亮等: "基于人脸检测和CAM SHIFT算法的人脸跟踪系统", 《微计算机应用》 * |
| 高建坡等: "以颜色和形状直方图为线索的粒子滤波人脸跟踪", 《中国图象图形学报》 * |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102833486B (en) * | 2012-07-05 | 2016-09-21 | 深圳泰山体育科技股份有限公司 | The method and device of face displaying ratio in a kind of real-time adjusting video images |
| CN102833486A (en) * | 2012-07-05 | 2012-12-19 | 深圳泰山在线科技有限公司 | Method and device for real-time adjusting face display scale in video image |
| CN103795851A (en) * | 2012-10-31 | 2014-05-14 | Lg电子株式会社 | Mobile terminal and method for controlling the same |
| CN103795851B (en) * | 2012-10-31 | 2018-10-19 | Lg电子株式会社 | Mobile terminal and its control method |
| CN107004116B (en) * | 2014-12-12 | 2018-09-21 | 北京市商汤科技开发有限公司 | Method and apparatus for predicting face's attribute |
| WO2016090522A1 (en) * | 2014-12-12 | 2016-06-16 | Xiaoou Tang | Method and apparatus for predicting face attributes |
| CN107004116A (en) * | 2014-12-12 | 2017-08-01 | 北京市商汤科技开发有限公司 | Method and apparatus for predicting face's attribute |
| CN106339072B (en) * | 2015-07-09 | 2019-04-23 | 中国科学院声学研究所 | A kind of distributed big data real time processing system and method based on left and right brain model |
| CN106339072A (en) * | 2015-07-09 | 2017-01-18 | 中国科学院声学研究所 | Distributed large data real-time processing system and method based on left and right brain model |
| US11003893B2 (en) | 2015-11-12 | 2021-05-11 | Advanced New Technologies Co., Ltd. | Face location tracking method, apparatus, and electronic device |
| US10713472B2 (en) | 2015-11-12 | 2020-07-14 | Alibaba Group Holding Limited | Face location tracking method, apparatus, and electronic device |
| CN106709932A (en) * | 2015-11-12 | 2017-05-24 | 阿里巴巴集团控股有限公司 | Face position tracking method and device and electronic equipment |
| US11423695B2 (en) | 2015-11-12 | 2022-08-23 | Advanced New Technologies Co., Ltd. | Face location tracking method, apparatus, and electronic device |
| CN105760826A (en) * | 2016-02-03 | 2016-07-13 | 歌尔声学股份有限公司 | Face tracking method and device and intelligent terminal. |
| US10747988B2 (en) | 2016-02-03 | 2020-08-18 | Goertek Inc | Method and device for face tracking and smart terminal |
| CN105760826B (en) * | 2016-02-03 | 2020-11-13 | 歌尔股份有限公司 | Face tracking method and device and intelligent terminal |
| CN105913028A (en) * | 2016-04-13 | 2016-08-31 | 华南师范大学 | Face tracking method and face tracking device based on face++ platform |
| CN108256532A (en) * | 2018-01-10 | 2018-07-06 | 武汉斗鱼网络科技有限公司 | Image processing method, device, electronic equipment and storage medium |
| CN113673461A (en) * | 2021-08-26 | 2021-11-19 | 深圳随锐云网科技有限公司 | Method and device for realizing selection of human face and human figure region based on 4K + AI |
| CN113673461B (en) * | 2021-08-26 | 2024-03-26 | 深圳随锐云网科技有限公司 | Method and device for realizing face and human shape area selection based on 4K+AI |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102054159B (en) | 2014-05-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102054159B (en) | Method and device for tracking human faces | |
| US8498444B2 (en) | Blob representation in video processing | |
| CN111345029B (en) | Target tracking method and device, movable platform and storage medium | |
| US20050180631A1 (en) | System and method for visual echo cancellation in a projector-camera-whiteboard system | |
| EP1845711A2 (en) | Electronic camera and image processing apparatus | |
| CN106709895A (en) | Image generating method and apparatus | |
| KR101491461B1 (en) | Method for recognizing object using covariance descriptor and apparatus thereof | |
| CN113302915B (en) | Method and system for generating a record of content appearing on a physical surface and captured on a video | |
| CN107589850A (en) | A kind of recognition methods of gesture moving direction and system | |
| CN109190522B (en) | Living body detection method based on infrared camera | |
| JP5484184B2 (en) | Image processing apparatus, image processing method, and program | |
| CN102682272A (en) | Electronic system, image correction method and computer program product thereof | |
| KR100845969B1 (en) | The Extraction method of moving object and the apparatus thereof | |
| KR102470873B1 (en) | Crop growth measurement device using image processing and method thereof | |
| CN110458753B (en) | Adaptive segmentation and undistorted unfolding system and method for panoramic girdle image | |
| CN104317385A (en) | Gesture identification method and system | |
| CN105844242A (en) | Method for detecting skin color in image | |
| US20180232192A1 (en) | System and Method for Visual Enhancement, Annotation and Broadcast of Physical Writing Surfaces | |
| CN108090428A (en) | A kind of face identification method and its system | |
| JP5677229B2 (en) | Video subtitle detection apparatus and program thereof | |
| CN112712054B (en) | Face wrinkle detection method | |
| CN107146258B (en) | Image salient region detection method | |
| JP2018018173A (en) | Image processing device, image processing method, and computer program | |
| CN108234770B (en) | Auxiliary makeup system, auxiliary makeup method and auxiliary makeup device | |
| CN117133032A (en) | Personnel identification and positioning method based on RGB-D image under face shielding condition |
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 |