CN106295587B - A kind of video interested region quick calibrating method - Google Patents
A kind of video interested region quick calibrating method Download PDFInfo
- Publication number
- CN106295587B CN106295587B CN201610675724.5A CN201610675724A CN106295587B CN 106295587 B CN106295587 B CN 106295587B CN 201610675724 A CN201610675724 A CN 201610675724A CN 106295587 B CN106295587 B CN 106295587B
- Authority
- CN
- China
- Prior art keywords
- value
- component
- pixel
- threshold
- macro block
- 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.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/161—Detection; Localisation; Normalisation
- G06V40/162—Detection; Localisation; Normalisation using pixel segmentation or colour matching
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/161—Detection; Localisation; Normalisation
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of video interested region quick calibrating method and device, the method comprises determining that the target frame image in video data；Determine the first component image data, second component image data in target frame image；It calculates the threshold value TH of the first component image data and calculates the difference threshold DTH of the second component image data of the first component image data and X times；According to the value of the first component of pixel and the value of second component determine the macro block in face or skin area in each macro block in the threshold value TH, the difference threshold DTH, the target frame image, the area-of-interest by the set of the macro block as the target frame image.The present invention provides a kind of methods of accurate and simple and quick calibration video interested region, and human face region is demarcated in dynamic video image, is particularly suitable for the limited terminal of processor ability and using situation.
Description
Technical field
The present invention relates to a kind of video data handling procedure more particularly to a kind of video interested region quick calibrating methods
With device.
Background technique
With internet, mobile Internet infrastructure constantly improve and the upgrading of cellular mobile communication networks is changed
Generation, people use visual telephone or video calling as the hand of remote communication exchange more and more in life and work
Section.But since network bandwidth is insufficient, network transmission load is excessive, video image resolution ratio is constantly promoted etc. it is many-sided because
The influence of element, traditional video coding technique not can guarantee the video image that user obtains high quality.Video coding technique is passed through
The development of many years how on the basis of metastable compression ratio, further obtains also gradually close to the limit of compression ratio and allows sight
Higher-quality video image in the person's of seeing subjective perception is one of the difficult point of Current video compression technical field.
Color space is also referred to as colour model (also known as color space or color system), its purposes is used under certain standards
Generally acceptable mode describes color.There are many kinds for color space, common to have RGB, YUV, HSI etc..RGB is (red green
It is blue) it is the space defined according to the color of eye recognition, it can indicate most of color.But it is bright because RGB model is by tone
Degree, saturation degree three are measured expression of putting together, are difficult to separate, it is difficult to digitized details adjustment is carried out, so in engineering and section
In grinding, RGB model is not used directly generally, but be after other colour models such as YUV or HSI carry out again by RGB model conversion
Continuous processing.
YUV colour model is that R, G of RGB model, B component are obtained luminance signal Y and two color difference by matrixing
Signal B-Y (i.e. U), R-Y (i.e. V), last transmitting terminal encodes three signals of brightness and color difference respectively, with same channel
It sends.Wherein " Y " indicates brightness (Luminance or Luma), that is, grayscale value；And " U " and " V " is indicated then
It is coloration (Chrominance or Chroma), effect is that description colors of image and saturation degree are divided for the color of specified pixel
It is not indicated with Cr and Cb.Wherein, Cr reflects the difference between RGB input signal RED sector and rgb signal brightness value, and
What Cb reflected is the difference between RGB input signal blue portion and rgb signal brightness value.The luminance signal Y of YUV colour model
It is separation with carrier chrominance signal U, V, is very suitable to the image procossings such as be enhanced, compressed.
HSI (Hue-Saturation-Intensity (Lightness), HSI or HSL) color model is joined with H, S, I tri-
Number description color characteristics, wherein H defines the wavelength of color, referred to as tone；S indicates gradation of color, referred to as saturation degree；I
Indicate intensity or brightness.When people observes a color body, the color of object is described with hue, saturation, intensity.Tone
It is the attribute (gilvous, orange or red) for describing pure color；Saturation degree provides a kind of pure color by the degree of the diluted degree of white light
Amount；Brightness is a subjective description, in fact, it cannot be measured, embodies colourless strength conception, and be to retouch
State the key parameter of color sensation.And intensity (gray scale) is monochrome image the most useful description, this amount be can measure and
It is easily explained.This model of proposition is then referred to as HSI (tone, saturation degree, intensity) color model, which can be in coloured silk
From the inner influence for eliminating strength component of the colour information of carrying (tone and saturation degree) in chromatic graph picture, opened so that HSI model becomes
The good tool of image processing method of the hair based on colour description, and this colored description is certainly however intuitive for people.
HSI model is U.S.'s chromatist's Munsell (H.A.Munseu) in proposition in 1915, it reflects the view of people
Feel the mode of system senses colour, color is perceived with three kinds of tone, saturation degree and intensity essential characteristic amounts.
Area-of-interest (ROI, Range of Interesting) Video coding is cope with the above problem at present effective
One of solution, since human eye has the characteristic of vision region of interest, i.e. human eye is interested to a sub-picture different zones
Degree is different, this means that human eye perceives degree also corresponding difference, therefore, base to the mass loss of piece image different zones
In the method for video coding of area-of-interest, this subjective characteristic of video image recipient can be made full use of, in image not
Different coding strategies is used with region, the subjectivity of image etc. can be effectively obviously improved while not increasing encoder bit rate
Experience quality.In order to realize this technology of interested area video coding, the coding and decoding video algorithm standard rules of current main-stream are mentioned
Corresponding suggestion implementation is supplied, such as International Organization for standardization/International Electrotechnical Commission (ISO/IEC) moving image is special
Family group -2 (MPEG-2) standard, the 10th partial higher video encoding standard of ISO/IEC mpeg-4 (MPEG-4)
(AVC), International Telecommunication Union (ITU) H.264 standard, be each provided with obtained in image specific region it is more high-quality than other regions
The mechanism of amount allows transmitting terminal to distinguish area-of-interest by analyzing the characteristic of image to be encoded, by area-of-interest
More coded-bits are distributed to retain more details, to obtain higher perceived quality.
And problem is extracted for general ROI, popular method is to establish computable visual attention location model, this is a kind of
It is related to vision physiological, cognitive psychological, memory mechanism, the complex model of image information etc. etc., it is difficult to quickly be marked
Method is determined, especially for dynamic image.Other processing specifically from dynamic image demarcate human face region method because
The property of privileged site in shape or face excessively to focus on face often introduces complicated algorithm, is not also suitable for
The Fast Calibration of human face region in dynamic image.The software encoder such as OPENH264 encoder for meeting H264 specification, due to calculating
Complexity of method itself can occupy a part of process resource of processor, thus the algorithm of wherein insertion human face region calibration must
It must be simple and quick and effective, and since the basic unit of compression processing in H264 is macro block (MB), human face region
Calibration should be the basic unit in the region with MB.
Demarcate about from image to human face region, there are many algorithms at present: Knowledge based engineering algorithm is utilized to face
The knowledge specified rule of priori is identified；Algorithm based on feature first finds the invariant feature of face and then carries out in the picture
This feature is verified in matching whether there is；There are also the algorithms based on statistics and probabilistic model.The colour of skin based on face various pieces
Feature is split processing, and consider each section shape feature and mutual positional relationship feature, this is a kind
In the algorithm based on feature, as other type algorithms, it also has sizable complexity, is not suitable for Fast Calibration.
The shortcomings that prior art, is mainly reflected in the complexity of calculating, and most of video related application is all dynamic vision
Frequently, there is comparable requirement of real-time, on the other hand, the equipment for carrying Video Applications, for example, common mobile terminal, processor energy
Power is limited, does not also allow too many computational complexity.
Summary of the invention
Processor energy is not suitable in order to solve video interested region calibration mode calculation complexity in the prior art
The problem of limited terminal of power or situation, the present invention provides a kind of video interested region quick calibrating method and devices.
The present invention provides a kind of video interested region quick calibrating methods, comprising:
Determine the target frame image in video data；
Determine the first component image data, second component image data in target frame image；
It calculates the threshold value TH of the first component image data and calculates the of the first component image data and X times
The difference threshold DTH of two component image data；
According in the threshold value TH, the difference threshold DTH, the target frame image first point of pixel in each macro block
The value of amount and the value of second component determine the macro block in face or skin area, using the set of the macro block as the target
The area-of-interest of frame image.
Above-mentioned video interested region quick calibrating method also has the following characteristics that
It is described according in the threshold value TH, the difference threshold DTH, the target frame image in each macro block pixel
The value of one component and the value of second component determine that the macro block in face or skin area includes:
When the value of the threshold value TH is greater than or equal to the first TH value, the value of the first component in each pixel of each macro block is determined
It is object pixel greater than the 2nd TH value and less than the pixel of the first TH value；
When the value of the threshold value TH is less than the first TH value and is greater than or equal to the 2nd TH value, each of each macro block is determined
The value of the first component is greater than the 2nd TH value and is less than the value and second component of the first TH value and first component in pixel
Value difference absolute value greater than the difference threshold DTH pixel be object pixel；
When the value of the threshold value TH is less than the 2nd TH value and is greater than or equal to the 3rd TH value, each of each macro block is determined
The value of the first component is greater than the exhausted of the 3rd TH value and the difference of the value of the value and second component of first component in pixel
Pixel to value greater than the difference threshold DTH is object pixel；
When the value of the threshold value TH is less than the 3rd TH value, determine that the value of the first component in each pixel of each macro block is big
In the 3rd TH value and it is less than the pixel of the first TH value for object pixel；
Determine object pixel number in each macro block be greater than the first ratio of number of pixels in the macro block macro block be in
The macro block of face or skin area；
Wherein, the first TH value, the 2nd TH value, the 3rd TH value successively reduce.
Above-mentioned video interested region quick calibrating method also has the following characteristics that
It is described according in the threshold value TH, the difference threshold DTH, the target frame image in each macro block pixel
The value of one component and the value of second component determine that the macro block in face or skin area includes:
Following operation is executed for each target frame image:
When the value of the threshold value TH is greater than or equal to the first TH value, the value of the first component in each pixel of each macro block is determined
It is object pixel greater than the 2nd TH value and less than the pixel of the first TH value；
The value of the threshold value TH determines each pixel of each macro block less than the first TH value and when being greater than or equal to the 2nd TH value
In the value of the first component be greater than first threshold and be less than the first TH value and the value of first component and the value of second component
Difference absolute value greater than the difference threshold DTH pixel be object pixel；First threshold is less than the 2nd TH value；
When the value of the threshold value TH is less than the 2nd TH value and is greater than or equal to the 3rd TH value, each of each macro block is determined
The value of the first component is greater than the threshold value TH and is greater than second threshold and the value of the value and second component of the first component in pixel
Pixel of the absolute value of difference greater than the difference threshold DTH is object pixel；
When the value of the threshold value TH is less than the 3rd TH value, determine that the value of the first component in each pixel of each macro block is big
In the threshold value TH and it is greater than third threshold value and is less than the pixel of the first TH value for object pixel；
Determine object pixel number in each macro block be greater than the first ratio of number of pixels in the macro block macro block be in
The macro block of face or skin area；
Wherein, the first TH value, the 2nd TH value, the 3rd TH value successively reduce；It is the first threshold, described
Second threshold, the third threshold value successively reduce and are all larger than the 3rd TH value and be less than the 2nd TH value.
Above-mentioned video interested region quick calibrating method also has the following characteristics that
The first component image data are the concentration component image data in the target frame image, the second component
Image data is the chromatic component image data in the target frame image；The value of the X be 1, the first TH value be 160,
The 2nd TH value is 145, the 3rd TH value is 120.
Above-mentioned video interested region quick calibrating method also has the following characteristics that
The first component image data are the concentration component image data in the target frame image, the second component
Image data is the chromatic component image data in the target frame image；The value of the X be 1, the first TH value be 160,
The 2nd TH value is 145, the 3rd TH value is 120, the first threshold is 140, the second threshold be 135, it is described
Third threshold value is 130.
Above-mentioned video interested region quick calibrating method also has the following characteristics that
The first component image data are the chrominance component image data in the target frame image, the second component
Image data is the saturation degree component image data in the target frame image；The value of the X is π, and the first TH value is 1.2
π, the 2nd TH value are 1.1 π, the 3rd TH value is π, the first threshold is 1.05 π, the second threshold be 1.02 π,
The third threshold value is 1.01 π.
Above-mentioned video interested region quick calibrating method also has the following characteristics that
First ratio is 1/10.
Above-mentioned video interested region quick calibrating method also has the following characteristics that
The threshold value TH for calculating the first component image data simultaneously calculates the first component image data and X times
The difference threshold DTH of second component image data includes: to calculate the first component image number according to maximum between-cluster variance method
According to threshold value TH and calculate the first component image data and X times second component image data difference threshold DTH.
Above-mentioned video interested region quick calibrating method also has the following characteristics that
Target frame image in the determining video data includes: to be chosen in the video data with same intervals frame number
Target frame image；Alternatively, according to the motion information of macro block determines target frame image in consecutive frame in the video data.
The present invention also provides a kind of video interested region fast calibration devices, comprising:
Target frame image determining module, for determining the target frame image in video data；
Component image data determining module, for determining the first component image data, second component figure in target frame image
As data；
Threshold calculation module, for calculating threshold value TH, the first component image number of the first component image data
According to the difference threshold DTH of the second component image data with X times；
Area-of-interest determining module, for according to the threshold value TH, the difference threshold DTH, the target frame image
In in each macro block the value of the first component of pixel and the value of second component determine the macro block in face or skin area, by institute
State area-of-interest of the set of macro block as the target frame image.
The present invention provides a kind of methods of accurate and simple and quick calibration video interested region, in dynamic vision
Human face region is demarcated in frequency, is particularly suitable for the limited terminal of processor ability and using situation.
Detailed description of the invention
The attached drawing for constituting a part of the invention is used to provide further understanding of the present invention, schematic reality of the invention
It applies example and its explanation is used to explain the present invention, do not constitute improper limitations of the present invention.In the accompanying drawings:
Fig. 1 is the flow chart of video interested region quick calibrating method in embodiment；
Fig. 2 is the structure chart of video interested region fast calibration device in embodiment.
Specific embodiment
Technical solution of the present invention is further illustrated with reference to the accompanying drawings and specific embodiments of the specification.It should be appreciated that this
Locate described specific embodiment to be used only for explaining the present invention, be not intended to limit the present invention.
Fig. 1 is the flow chart of video interested region quick calibrating method in embodiment, and the method includes:
Step 101, the target frame image in video data is determined；
Step 102, the first component image data, second component image data in target frame image are determined；
Step 103, threshold value TH, the first component image data and X times of the second component of the first component image data are calculated
The difference threshold DTH of image data；
Step 104, according to the first component of pixel in each macro block in threshold value TH, difference threshold DTH, target frame image
Value and the value of second component determine the macro block in face or skin area, by the set of the macro block as target frame image
Area-of-interest.
Wherein, the mode for the target frame image in video data being determined in step 101 includes following two: with same intervals
Target frame image in frame number selecting video data；Alternatively, according to the motion information of macro block determines in consecutive frame in video data
Target frame image when the motion range of macro block is smaller in consecutive frame, determines target frame image with lesser interval frame number, in phase
When the motion range of macro block is larger in adjacent frame, target frame image is determined with biggish interval frame number.The present invention is for processing capacity
Poor equipment, it is easy to by abandoning the performance of sub-fraction human face region calibration, equivalent promotion calculation proposed by the present invention
The speed of method, it is specific as follows: using the correlation between continuous picture frame, the frame number greater than 1 to can be used, such as with 2 or 3
Frame does face region labeling for the period rather than every 1 frame is all used as target frame, and the frame in the period thinks that human face region is constant.It is regarding
In the case that the picture material variation of frequency stream is unhappy, the performance for hardly losing calibration is done so.
In step 103, the threshold value TH and the first component map of the first component image data are calculated according to maximum between-cluster variance
As the threshold value DTH of data and the difference of X times of second component image data.
Below by using two of concentration component V and chromatic component U in video image to illustrate realization of the invention
Method.When the picture format of target frame image in step 102 is not yuv format, need that target frame image is first converted to YUV figure
As format executes operations described below again.
Implementation method one
The Fast Calibration of ROI is carried out in this implementation method two using yuv space.Wherein, the first component image data are mesh
The concentration component V image data in frame image are marked, second component image data is the chromatic component U picture number in target frame image
According to.
Following operation is executed for each target frame image:
When the value of TH is greater than or equal to the first TH value, determine that the value of the first component in each pixel of each macro block is greater than second
TH value and less than the pixel of the first TH value be object pixel；
The value of TH determines in each pixel of each macro block first point less than the first TH value and when being greater than or equal to the 2nd TH value
The value of amount is greater than the 2nd TH value and big less than the absolute value of the first TH value and the difference of the value of the value and second component of the first component
In difference threshold DTH pixel be object pixel；
The value of TH determines in each pixel of each macro block first point less than the 2nd TH value and when being greater than or equal to the 3rd TH value
The value of amount is greater than the 3rd TH value and the absolute value of the difference of the value of the value and second component of the first component is greater than difference threshold DTH's
Pixel is object pixel；
When the value of TH is less than the 3rd TH value, determine the first component in each pixel of each macro block value be greater than the 3rd TH value and
Pixel less than the first TH value is object pixel；
The macro block for determining that object pixel number is greater than the first ratio of number of pixels in macro block in each macro block is in face
Or the macro block of skin area.
Wherein, the first TH value, the 2nd TH value, the 3rd TH value successively reduce.
It is as follows for the specific example of the method one:
The threshold value of concentration component V image data in target frame image is V_TH, the first component V image data and second point
The difference threshold for measuring U image data is DVU_TH.It is the 145, the 3rd TH value is 120 that first TH value, which is the 160, the 2nd TH value,.
Following operation is executed for each target frame image:
When V_TH>=160, determine that the pixel of V>145 and V<160 is object pixel in each pixel of each macro block；
V_TH>=145 and when V_TH<160, determine in each pixel of each macro block V>145 and V<160 and | V-U |>
The pixel of DVU_TH is object pixel；
V_TH>=120 and when V_TH<145, determine in each pixel of each macro block V>120 and | V-U |>DVU_TH's
Pixel is object pixel；
When V_TH<120, determine that the pixel of V>120 and V<160 is object pixel in each pixel of each macro block.
Implementation method two
The Fast Calibration of ROI is carried out in this implementation method two using yuv space.Wherein, the first component image data are mesh
The concentration component V image data in frame image are marked, second component image data is the chromatic component U picture number in target frame image
According to.
Following operation is executed for each target frame image:
When the value of threshold value TH is greater than or equal to the first TH value, determine that the value of the first component in each pixel of each macro block is greater than
2nd TH value and less than the pixel of the first TH value be object pixel；
The value of threshold value TH is less than the first TH value and when being greater than or equal to the 2nd TH value, determines in each pixel of each macro block the
The value of one component is greater than first threshold and absolute less than the difference of the first TH value and the value of the value and second component of the first component
Pixel of the value greater than difference threshold DTH is object pixel；First threshold is less than the 2nd TH value；
The value of threshold value TH is less than the 2nd TH value and when being greater than or equal to the 3rd TH value, determines in each pixel of each macro block the
The value of one component is greater than threshold value TH and is greater than the absolute value of second threshold and the difference of the value of the value and second component of the first component
Pixel greater than difference threshold DTH is object pixel；
When the value of threshold value TH is less than the 3rd TH value, determine that the value of the first component in each pixel of each macro block is greater than threshold
Value TH and be greater than third threshold value and less than the pixel of the first TH value be object pixel；
The macro block for determining that object pixel number is greater than the first ratio of number of pixels in macro block in each macro block is in face
Or the macro block of skin area.
Wherein, the first TH value, the 2nd TH value, the 3rd TH value successively reduce.First threshold, second threshold, third threshold value according to
Secondary reduction and it is all larger than the 3rd TH value and less than the 2nd TH value.
It is as follows for the specific example of the method two:
The threshold value of concentration component V image data in target frame image is V_TH, the first component V image data and second point
The difference threshold for measuring U image data is DVU_TH.It is the 145, the 3rd TH value is 120, first that first TH value, which is the 160, the 2nd TH value,
Threshold value is 140, second threshold 135, third threshold value are 130.
Following operation is executed for each target frame image:
When V_TH>=160, determine that the pixel of V>145 and V<160 is object pixel in each pixel of each macro block；
V_TH>=145 and when V_TH<160, determine in each pixel of each macro block V>140 and V<160 and | V-U |>
The pixel of DVU_TH is object pixel；
V_TH>=120 and when V_TH<145, determine in each pixel of each macro block V>V_TH and V>135 and | V-U |>
The pixel of DVU_TH is object pixel；
When V_TH<120, determine that the pixel of V>V_TH and V>130 and V<160 is target picture in each pixel of each macro block
Element.
The macro block for determining that object pixel number is greater than the first ratio of number of pixels in macro block in each macro block is in face
Or the macro block of skin area.
Implementation method three
The Fast Calibration of ROI is carried out in this implementation method three using the space HSI.Wherein, the first component image data are mesh
The chrominance component H in frame image is marked, second component image data is the saturation degree component S in target frame image.
Following operation is executed for each target frame image:
When the value of threshold value TH is greater than or equal to the first TH value, determine that the value of the first component in each pixel of each macro block is greater than
First TH value and less than the pixel of the first TH value be object pixel；
The value of threshold value TH is less than the first TH value and when being greater than or equal to the 2nd TH value, determines in each pixel of each macro block the
The value of one component is greater than first threshold and absolute less than the difference of the first TH value and the value of the value and second component of the first component
Pixel of the value greater than threshold value DTH is object pixel；First threshold is less than the 2nd TH value；
The value of threshold value TH is less than the 2nd TH value and when being greater than or equal to the 3rd TH value, determines in each pixel of each macro block the
The value of one component is greater than threshold value TH value and is greater than the absolute of second threshold and the difference of the value of the value and second component of the first component
Pixel of the value greater than difference binarization threshold DTH is object pixel；
When the value of threshold value TH is less than the 3rd TH value, determine that the value of the first component in each pixel of each macro block is greater than threshold value TH
It is worth and is greater than third threshold value and is object pixel less than the pixel of the first TH value；
Wherein, the first TH value, the 2nd TH value, the 3rd TH value successively reduce；
First threshold, second threshold, third threshold value successively reduce.
The macro block for determining that object pixel number is greater than the first ratio of number of pixels in macro block in each macro block is in face
Or the macro block of skin area.
It is as follows for the specific example of this implementation method three:
The threshold value of chrominance component H image data in target frame image is H_TH, the first component H image data and second point
The threshold value for measuring π times of difference of S image data is DHS_TH.First TH value is 1.2 π, the 2nd TH value is 1.1 π, the 3rd TH value
It is 1.05 π for π, first threshold, second threshold is 1.02 π, third threshold value is 1.01 π.
Following operation is executed for each target frame image:
When H_TH>=1.2* π, determine that the pixel of the π of H>1.1 and the π of H<1.2 are object pixel in each pixel of each macro block；
The π of H_TH>=1.1 and when H_TH<1.2 π, determine in each pixel of each macro block the π of H>1.1 and the π of H<1.2 and | H-
S* π | the pixel of > DHS_TH is object pixel；
H_TH>=π and when H_TH<1.1 π, determine in each pixel of each macro block H>H_TH and the π of H>1.02 and | H-S*
π | the pixel of > DHS_TH is object pixel；
When H_TH<π, determine that the pixel of V>H_TH and the π of V>1.01 and the π of V<1.2 are target picture in each pixel of each macro block
Element.
The macro block for determining that object pixel number is greater than the first ratio of number of pixels in macro block in each macro block is in face
Or the macro block of skin area.
In the method, the first ratio is preferably but not limited to 1/10, and macro block determines target picture in each macro block when being 16X16 macro block
Macro block of the plain number greater than 16 is the macro block in face or skin area.
The carry out human face region calibration in the picture of complexion model and maximum variance between clusters (OSTU) is based in this method
Fast method, different from the method being split based on face various pieces features of skin colors, it is without the concern for geometry
And the positional relationship between these shapes, and several threshold values only are obtained after simple statistics calculates, then pass through inspection
Look into the relationship between the UV value of the pixel in MB and these threshold values, can one MB of fairly accurate calibration whether in human face region
Or in the skin area of exposure.
The present invention is not limited to the mode of the HS component of the mode and HIS space using the UV component of yuv space, can be with
It is the component in rgb space, can also be the component in other picture formats or other color spaces, use other picture formats
Or other color spaces component when respective threshold need to be changed accordingly according to picture format or color space.
The present invention also provides the video interested region of the video interested region quick calibrating method corresponding to Fig. 1 is fast
Speed variator.Fig. 2 is video interested region fast calibration device in embodiment, with reference to Fig. 2, video interested region Fast Calibration
Device includes:
Target frame image determining module 201, for determining the target frame image in video data；
Component image data determining module 202, for determining the first component image data, second component in target frame image
Image data；
Threshold calculation module 203, for calculating threshold value TH, the first component image data and the X of the first component image data
The difference threshold DTH of second component image data again；
Area-of-interest determining module 204, for according to each macro block in threshold value TH, difference threshold DTH, target frame image
The value of first component of middle pixel and the value of second component determine the macro block in face or skin area, by the collection of the macro block
Cooperation is the area-of-interest of target frame image.
The specific implementation of the function of above-mentioned each module is identical as mode described in the above method, is not repeated herein
Description.
Method of the invention is simple and quick and accurate and effective, can be all with software realization, and implementation when is very flexible, several
There is no cost, is very suitable for using in video image processing especially internet video application: once human face region quilt
Calibration, the subsequent specially treated for the region are known as possibility.Such as: when using H.264 encoding, face area can be promoted
The quality in domain, while the quality of the appropriate other parts for reducing image, it is overall to keep encoding rate constant, in this way, even if in network
The figure that in the case where Bandwidth-Constrained, after compressed bit stream is decoded, can also obtain face clearly image, rather than obscure all over
Picture.
Those of ordinary skill in the art will appreciate that all or part of the steps in the above method can be instructed by program
Related hardware is completed, and described program can store in computer readable storage medium, such as read-only memory, disk or CD
Deng.Optionally, one or more integrated circuits also can be used to realize, accordingly in all or part of the steps of above-described embodiment
Ground, each module/unit in above-described embodiment can take the form of hardware realization, can also use the shape of software function module
Formula is realized.The present invention is not limited to the combinations of the hardware and software of any particular form.
Descriptions above can combine implementation individually or in various ways, and these variants all exist
Within protection scope of the present invention.
It should be noted that, in this document, the terms "include", "comprise" or its any other variant are intended to non-row
His property includes, so that including the article of a series of elements or equipment not only includes those elements, but also including not having
There is the other element being expressly recited, or further includes for this article or the intrinsic element of equipment.Do not limiting more
In the case where system, the element that is limited by sentence " including ... ", it is not excluded that in the article or equipment for including the element
There is also other identical elements.
The above examples are only used to illustrate the technical scheme of the present invention and are not limiting, reference only to preferred embodiment to this hair
It is bright to be described in detail.Those skilled in the art should understand that can modify to technical solution of the present invention
Or equivalent replacement should all cover in claim model of the invention without departing from the spirit and scope of the technical solution of the present invention
In enclosing.
Claims (7)
1. a kind of video interested region quick calibrating method characterized by comprising
Determine the target frame image in video data；
Determine the first component image data, second component image data in target frame image；
It calculates the threshold value TH of the first component image data and calculates the first component image data and X times of second component
The difference threshold DTH of image data；
According to the first component of pixel in each macro block in the threshold value TH, the difference threshold DTH, the target frame image
Value and the value of second component determine the macro block in face or skin area, using the set of the macro block as the target frame figure
The area-of-interest of picture；
It is described according in the threshold value TH, the difference threshold DTH, the target frame image first point of pixel in each macro block
The value of amount and the value of second component determine that the macro block in face or skin area includes:
When the value of the threshold value TH is greater than or equal to the first TH value, determine that the value of the first component in each pixel of each macro block is greater than
2nd TH value and less than the pixel of the first TH value be object pixel；
When the value of the threshold value TH is less than the first TH value and is greater than or equal to the 2nd TH value, each pixel of each macro block is determined
In the value of the first component be greater than the 2nd TH value and be less than the first TH value and the value of first component and the value of second component
Difference absolute value greater than the difference threshold DTH pixel be object pixel；
When the value of the threshold value TH is less than the 2nd TH value and is greater than or equal to the 3rd TH value, each pixel of each macro block is determined
In the value of the first component be greater than the absolute value of the 3rd TH value and the difference of the value of the value and second component of first component
Pixel greater than the difference threshold DTH is object pixel；
When the value of the threshold value TH is less than the 3rd TH value, determine the value of the first component in each pixel of each macro block greater than the
Three TH values and the pixel for being less than the first TH value are object pixel；
The macro block for determining that object pixel number is greater than the first ratio of number of pixels in the macro block in each macro block is in face
Or the macro block of skin area；
Wherein, the first TH value, the 2nd TH value, the 3rd TH value successively reduce；
Alternatively,
It is described according in the threshold value TH, the difference threshold DTH, the target frame image first point of pixel in each macro block
The value of amount and the value of second component determine that the macro block in face or skin area includes:
Following operation is executed for each target frame image:
When the value of the threshold value TH is greater than or equal to the first TH value, determine that the value of the first component in each pixel of each macro block is greater than
2nd TH value and less than the pixel of the first TH value be object pixel；
The value of the threshold value TH is less than the first TH value and when being greater than or equal to the 2nd TH value, determines in each pixel of each macro block the
The value of one component is greater than first threshold and is less than the difference of the first TH value and the value of the value and second component of first component
Pixel of the absolute value of value greater than the difference threshold DTH is object pixel；First threshold is less than the 2nd TH value；
When the value of the threshold value TH is less than the 2nd TH value and is greater than or equal to the 3rd TH value, each pixel of each macro block is determined
In the value of the first component be greater than the threshold value TH and be greater than the difference of second threshold and the value of the value and second component of the first component
Absolute value greater than the difference threshold DTH pixel be object pixel；
When the value of the threshold value TH is less than the 3rd TH value, determine that the value of the first component in each pixel of each macro block is greater than institute
Stating threshold value TH and being greater than third threshold value and be less than the pixel of the first TH value is object pixel；
The macro block for determining that object pixel number is greater than the first ratio of number of pixels in the macro block in each macro block is in face
Or the macro block of skin area；
Wherein, the first TH value, the 2nd TH value, the 3rd TH value successively reduce；The first threshold, described second
Threshold value, the third threshold value successively reduce and are all larger than the 3rd TH value and be less than the 2nd TH value.
2. video interested region quick calibrating method as described in claim 1, which is characterized in that
The first component image data are the concentration component image data in the target frame image, the second component image
Data are the chromatic component image datas in the target frame image；The value of the X be 1, the first TH value be 160, it is described
2nd TH value is 145, the 3rd TH value is 120.
3. video interested region quick calibrating method as described in claim 1, which is characterized in that
The first component image data are the concentration component image data in the target frame image, the second component image
Data are the chromatic component image datas in the target frame image；The value of the X be 1, the first TH value be 160, it is described
2nd TH value is 145, the 3rd TH value is 120, the first threshold is 140, and the second threshold is 135, the third
Threshold value is 130.
4. video interested region quick calibrating method as described in claim 1, which is characterized in that
The first component image data are the chrominance component image data in the target frame image, the second component image
Data are the saturation degree component image data in the target frame image；The value of the X is π, and the first TH value is 1.2 π, institute
State that the 2nd TH value is 1.1 π, the 3rd TH value is π, the first threshold is 1.05 π, the second threshold is 1.02 π, described
Third threshold value is 1.01 π.
5. video interested region quick calibrating method as described in claim 1, which is characterized in that
First ratio is 1/10.
6. video interested region quick calibrating method as described in claim 1, which is characterized in that
The threshold value TH for calculating the first component image data simultaneously calculates the first component image data and the second of X times
The difference threshold DTH of component image data includes: to calculate the first component image data according to maximum between-cluster variance method
Threshold value TH and calculate the first component image data and X times second component image data difference threshold DTH.
7. video interested region quick calibrating method as described in claim 1, which is characterized in that
Target frame image in the determining video data includes: the target chosen in the video data with same intervals frame number
Frame image；Alternatively, according to the motion information of macro block determines target frame image in consecutive frame in the video data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610675724.5A CN106295587B (en) | 2016-08-16 | 2016-08-16 | A kind of video interested region quick calibrating method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610675724.5A CN106295587B (en) | 2016-08-16 | 2016-08-16 | A kind of video interested region quick calibrating method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106295587A CN106295587A (en) | 2017-01-04 |
| CN106295587B true CN106295587B (en) | 2019-05-31 |
Family
ID=57678217
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610675724.5A Active CN106295587B (en) | 2016-08-16 | 2016-08-16 | A kind of video interested region quick calibrating method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106295587B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108737821B (en) * | 2018-04-25 | 2020-09-04 | 中国人民解放军军事科学院军事医学研究院 | Video interest area quick preselection method and system based on multi-channel shallow feature |
| CN110689014B (en) * | 2019-10-14 | 2022-02-11 | 腾讯科技（深圳）有限公司 | Method and device for detecting region of interest, electronic equipment and readable storage medium |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP4657934B2 (en) * | 2006-01-23 | 2011-03-23 | 富士フイルム株式会社 | Face detection method, apparatus and program |
| CN101882315B (en) * | 2009-05-04 | 2015-11-25 | 海信集团有限公司 | The detection method of area of skin color |
| CN102663682B (en) * | 2012-03-13 | 2015-06-03 | 华中科技大学 | Adaptive image enhancement method based on interesting area |
| CN103514611B (en) * | 2012-06-26 | 2016-08-03 | 北京新媒传信科技有限公司 | The extracting method of a kind of area of skin color and device |
| CN104394344B (en) * | 2014-11-26 | 2018-02-02 | 北京智谷睿拓技术服务有限公司 | Video frame selecting method and equipment |
-
2016
- 2016-08-16 CN CN201610675724.5A patent/CN106295587B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN106295587A (en) | 2017-01-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102884536B (en) | The colour of skin and feature detection for video conference compression | |
| KR101703965B1 (en) | Method of processing of a compressed image into a gamut mapped image using spatial frequency analysis | |
| ES2386872T3 (en) | Method and system to improve chrominance information of a compressed image | |
| CN107203974A (en) | The methods, devices and systems of HDR HDR to the HDR tones mapping of extension | |
| CN106357955B (en) | The scalable system for being used to control color management of metadata grade including variation | |
| WO2019092463A1 (en) | Video image processing | |
| CN102812705B (en) | image encoder and image decoder | |
| CN108293124A (en) | The coding of picture in video | |
| CN104581103B (en) | A kind of image processing method and device | |
| JP2019512953A (en) | Saturation specification for dynamic range mapping | |
| CN107615761A (en) | Processes pixel and coding | |
| CN101317185B (en) | Video sensor-based automatic region-of-interest detection | |
| TWI573443B (en) | Video coding using a saliency map | |
| US20170094281A1 (en) | Compressing high dynamic range images | |
| JP2009508427A (en) | Image enhancement and compression | |
| CN106295587B (en) | A kind of video interested region quick calibrating method | |
| CA2825927A1 (en) | Color conversion based on an hvs model | |
| CN103180864A (en) | Method for efficient representation and processing of color pixel data in digital pathology images | |
| WO2014018050A1 (en) | Techniques for Video Compression | |
| JP3387071B2 (en) | Image identification apparatus and method | |
| Lee et al. | Contrast-preserved chroma enhancement technique using YCbCr color space | |
| CN108205795A (en) | Face image processing process and system during a kind of live streaming | |
| Lu et al. | Compression efficiency improvement over HEVC main 10 profile for HDR and WCG content | |
| KR100406535B1 (en) | Method of image compression and device for the same | |
| Yoo et al. | Colour naturalness metric for evaluating image quality of mobile displays |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20210714 Address after: 100044 1320-2, 13th floor, building 1, yard 59, gaoliangqiaoxie street, Haidian District, Beijing Patentee after: Beijing Anxin Zhitong Technology Co.,Ltd. Address before: 100192 room b-2d203, Northern Territory, Dongsheng Science Park, 66 xixiaokou Road, Haidian District, Beijing Patentee before: YEECALL NETWORK Ltd. |