CN108764254A - A kind of image characteristic point describes method - Google Patents
A kind of image characteristic point describes method Download PDFInfo
- Publication number
- CN108764254A CN108764254A CN201810488159.0A CN201810488159A CN108764254A CN 108764254 A CN108764254 A CN 108764254A CN 201810488159 A CN201810488159 A CN 201810488159A CN 108764254 A CN108764254 A CN 108764254A
- Authority
- CN
- China
- Prior art keywords
- small echo
- video
- operation values
- space
- characteristic point
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of image characteristic points to describe method, since in the method, description of space-time characteristic point contains the half-tone information and movable information of video image；This method rotates small echo operation values simultaneously, ensure that the rotational invariance of description；And the small echo response that the small echo response that description is utilized is not only space-time characteristic point is obtained, further include the small echo response of each pixel in the region centered on space-time characteristic point.Based on above 3 points, what disclosed method obtains through the invention description can be more precisely characterizes video image, to avoid the problem that unilateral being characterized to video.
Description
Technical field
The present invention relates to technical field of video processing, and method is described more specifically to a kind of image characteristic point.
Background technology
In existing video characterizing method, a video is often resolved into the image of multiple single frames, by obtaining video figure
Half-tone information as upper pixel and the half-tone information according to pixel realize the characterization to video, but the letter that a video is included
Breath never only has the half-tone information of pixel, further includes the pixel interdependence between video time sequence, therefore the characterization video
Method be unilateral, video is not characterized according to the temporal information for including in video.
Characteristic point in the extraction image pixel used with reference to the prior art, and characteristic point is described to realization table
The method for levying image, it may be considered that include the space-time characteristic point of spatial information and temporal information in extraction video, and to the space-time
Characteristic point is described to realize to the characterization of video, and how to be described and be realized pair according to the space-time characteristic point of video
The characterization of video is then urgently to be resolved hurrily.
Invention content
The main purpose of the present invention is to provide a kind of image characteristic points to describe method, it is intended to solve existing video characterization
The problem of partial information that method uses only video carries out unilateral characterization to video.
To achieve the above object, a kind of image characteristic point of present invention offer describes method, and this method includes：
Step A1, cylindrical region is constructed centered on the space-time characteristic point on the integral video of video image, and will accumulated
Divide video to carry out small echo operation in the predeterminable area of cylindrical region, obtains integral video in each predeterminable area of cylindrical region
Small echo operation values, the small echo operation values of more each predeterminable area, by the direction corresponding to the maximum predeterminable area of small echo operation values
It is determined as the principal direction θ of space-time characteristic point；Cylindrical region at least has there are one predeterminable area, and integral video is to be based on video figure
The half-tone information of each pixel and movable information obtain as in；
Step A2, neighborhood is chosen centered on the space-time characteristic point on the integral video of video image, and video will be integrated
Small echo operation is carried out in the subregion of neighborhood, obtains integral video in the small echo operation values of all subregion, neighborhood is at least with one
Sub-regions；
Step A3, the small echo operation values of all subregion are rotated using the principal direction θ of space-time characteristic point, and based on rotation
The response vector of all subregion of space-time characteristic point is calculated in small echo operation values after turning；
Step A4, the response vector of all subregion of statistics space-time characteristic point, and carried out using the response vector after statistics
It calculates, obtains description of space-time characteristic point.
Optionally, step A1 includes the following steps：
Step A11, cylindrical region is constructed centered on the space-time characteristic point on the integral video of video image, with default
The three-dimensional sector region of size is slided in cylindrical region according to default sliding type, and integral video is fanned in three-dimensional
Shape region carry out small echo operation, obtain integral video three-dimensional sector region small echo operation values；
Three-dimensional sector region is slided according to default sliding type, will obtain at least one three-dimensional fan in cylindrical region
Shape region；
Step A12, each three-dimensional fan section is calculated in the small echo operation values of each three-dimensional sector region using integral video
The characteristic vector in domain, characteristic vector include modulus value and direction；
Step A13, the characteristic vector of each three-dimensional sector region of comparison, by the maximum three-dimensional sector region characteristic vector of modulus value
Corresponding direction is determined as the principal direction θ of space-time characteristic point.
Optionally, integral video includes F in the small echo operation values of subregionx、FyAnd Ft, Fx、Fy、FtIndicate that integral regards respectively
Frequency is in subregion x, y, the small echo operation values in the directions t, then step A3 includes：
According to formula Fx'=- Fx×sin(θ)+Fy× cos (θ), Fy'=Fx×cos(θ)+Fy× sin (θ), Ft'=Ft,
To Fx、FyAnd FtIt is rotated, obtains integral video in subregion x, y, the postrotational small echo operation values F in the directions tx'、Fy' and
Ft'；
Statistics integral video is in the postrotational small echo operation values F of subregion all directionsx'、Fy' and Ft', obtain ∑ Fx'、∑|
Fx'|、∑Fy'、∑|Fx'|、∑Ft'、∑|Ft'|；
According to Vr=[∑ Fx',∑|Fx'|,∑Fy',∑|Fy'|,∑Ft',∑|Ft' |] that space-time characteristic point is calculated is each
The response vector V of subregionr, response vector VrIndicate the response vector of subregion r.
Advantageous effect
The description method of video image space-time characteristic point provided by the present invention includes above-mentioned step A1~A4, due to when
Description of empty characteristic point is obtained based on integral video, and integrates the half-tone information that video comprehensively contains video image
And movable information, therefore description method provided by the invention comprehensively uses video image information, description through the invention
Method can be to avoid unilateral the problem of being characterized to video；Simultaneously as description method also carries out small echo operation values
Rotation is further realized by description with rotational invariance to regarding so ensure that the rotational invariance of description
The characterization of frequency image can avoid unilateral the problem of being characterized to video；On the other hand, description of the invention method be to
Region centered on space-time characteristic point carries out small echo operation and obtains small echo response, is based on small echo response later and is worth to description
Son, by obtaining the small echo response that the small echo response that is utilized of description is not only space-time characteristic point, further include with when
The small echo response of each pixel in region centered on empty characteristic point, thus obtain description son can be more precisely to space-time spy
Sign point characterize and then is characterized to video image, to avoid the problem that unilateral being characterized to video.
Description of the drawings
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technology description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of invention for those skilled in the art without creative efforts, can also basis
These attached drawings obtain other attached drawings.
Fig. 1 is the basic flow chart that the video image space-time characteristic point that first embodiment of the invention provides describes method；
Fig. 2 is the structural schematic diagram for the description device that second embodiment of the invention provides.
Specific implementation mode
In order to make the invention's purpose, features and advantages of the invention more obvious and easy to understand, below in conjunction with the present invention
Attached drawing in embodiment, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described reality
It is only a part of the embodiment of the present invention to apply example, and not all embodiments.Based on the embodiments of the present invention, people in the art
The every other embodiment that member is obtained without making creative work, shall fall within the protection scope of the present invention.
First embodiment
Fig. 1 is the basic flow chart of the description method of video image space-time characteristic point provided in this embodiment, this method packet
It includes：
S101, its principal direction is determined according to the space-time characteristic point on the integral video of video image.
The S101 the specific steps are：It is constructed centered on the space-time characteristic point on the integral video of video image cylindrical
Region, and predeterminable area of the video in cylindrical region will be integrated and carry out small echo operation, integral video is obtained in cylindrical region
The small echo operation values of each predeterminable area, the small echo operation values of more each predeterminable area, by the maximum predeterminable area of small echo operation values
Corresponding direction is determined as the principal direction θ of space-time characteristic point.
It is to be appreciated that above-mentioned cylindrical region at least has there are one predeterminable area, integral video is to be based on video
The half-tone information of each pixel and movable information obtain in image.Integral video is introduced herein：Using gamma function f
(p) and movement function vpThe half-tone information and movable information of pixel in a video can be characterized, therefore, by the position of each pixel
Coordinate brings above-mentioned gamma function and movement function into, you can obtains the half-tone information and movable information of each pixel, is based on each picture
The half-tone information and movable information of vegetarian refreshments can obtain gray scale movement function f'(p), and it is based on gray scale movement function f'(p) can be into
One step determines integral video V (p), it will be understood that there are three variables for gray scale movement function and integral video tool, to sit
Parameter is the three-dimensional coordinate function of x, y, t.Specifically, based on the half-tone information f (p) of each pixel p in video image and movement
Information vp, according toF'(p)=f (p)+vp, the integral video V (p) of video image can be calculated,
Middle S0Indicate the cube formed from origin o to p, p ∈ S0。
It is to be understood that integral video is a kind of three-dimensional coordinate function for describing video image, by the three-dimensional coordinate letter
Number, which is put into three-dimensional coordinate, will obtain " three-dimensional " integral video corresponding to the function, you can to understand that integral video is by multiple pictures
The three-dimensional structure that plain cube is constituted will integrate the concept of video to follow-up with " three-dimensional " hereinafter for ease of illustrating
The step of be introduced, but the concept its integral video can't be generated it is further limit, meanwhile, be hereinafter previously mentioned
Integral video, specifically three-dimensional coordinate function or above-mentioned " three-dimensional " integrate video, then needs according to mentioning integral video
Depending on situation, no longer it is explained hereinafter.
Step S101 is further explained based on the above content：Circle is constructed on video centered on space-time characteristic point to integrate
Predeterminable area is chosen on the cylindrical region in column regions, it is possible to understand that the predeterminable area is also the solid region of three-dimensional, it
Integral video is subjected to small echo operation in the predeterminable area afterwards, integral video is obtained and corresponds to predeterminable area in the cylindrical region
Small echo operation values will further integrate other predeterminable areas of video in cylindrical region and carry out small echo operation, obtain each default
The small echo response in region, it is to be understood that integral video is subjected to small echo operation in predeterminable area and is referred to the predeterminable area
Interior integral video is responded using Haar small echos (one kind in small echo is simplest orthonomalization small echo) template
Value calculates.After obtaining the small echo response of each predeterminable area, the small echo operation values of more each predeterminable area, by small echo operation values
Direction corresponding to maximum predeterminable area is determined as the principal direction of space-time characteristic point, it is to be understood that " predeterminable area herein
Small echo operation values " refer to small echo response corresponding to integral video image vegetarian refreshments all in predeterminable area set, this is pre-
If region has direction, it can determine that small echo response value set is maximum by comparing the small echo response value set of each predeterminable area
Predeterminable area, the direction for further having the predeterminable area is as the direction of space-time characteristic point.
In the other example of the present embodiment, such as the first example is (it is understood that " Z examples " herein is only
In order to refer to some example, only have this several example being previously mentioned without limiting the present embodiment) in, step S101 includes：
S1011, cylindrical region is constructed centered on the space-time characteristic point on the integral video of video image, with default big
Small three-dimensional sector region is slided in cylindrical region according to default sliding type, and will integrate video three-dimensional fan-shaped
Region carry out small echo operation, obtain integral video three-dimensional sector region small echo operation values.
It is to be appreciated that three-dimensional sector region is slided according to default sliding type, will be obtained in cylindrical region
At least one three-dimensional sector region, obtained three-dimensional sector region is each predeterminable area being previously mentioned in above-described embodiment.
It illustrates herein and cylindrical region is introduced：Centered on space-time characteristic point, radius is 6 σx, a height of σtCylinder
Response calculating is carried out using Haar small echo templates to integral video in shape region, then cylindrical region G is represented by：
Wherein (x0,y0,t0) be space-time characteristic point coordinate, σxAnd σtRespectively space-time characteristic point is in spatial domain and time-domain
Scale.
In the first exemplary one kind, three-dimensional sector region is the three-dimensional sector region that subtended angle is π/3, is preset
Sliding type is to carry out rotational slide in cylindrical region with 0.2 radian of step-length.
S1012, using integral video in the small echo operation values of each three-dimensional sector region, each three-dimensional sector region is calculated
Characteristic vector.Characteristic vector at this includes modulus value and direction.
In the case that this first it is exemplary another, step S1012 further includes：It is regarded using three-dimensional fan section domain integral
Frequently the small echo operation values of each pixel, according to Mould is calculated
Value mwWith direction θw, by (mw,θw) characteristic vector as three-dimensional sector region, used three when w indicates to calculate characteristic vector
Tie up sector region.
It is to be appreciated that integral video is made of at least one pixel, integral video is in the small of three-dimensional sector region
Wave operation values include the small echo operation values of three-dimensional each pixel of fan section domain integral video.In epimere content, video is integrated
The small echo operation values of pixel include Ex、Ey, and Ex、EyIndicate the pixel of integral video in x, the small echo operation in the directions y respectively
Value.
The characteristic vector of S1013, each three-dimensional sector region of comparison, by the maximum three-dimensional sector region characteristic vector institute of modulus value
Corresponding direction is determined as the principal direction of space-time characteristic point.
Based on above-mentioned first another exemplary situation, principal direction can be according to formula θ=θw|max{mwAcquire.
S102, small echo operation values of the integral video in all subregion of space-time characteristic vertex neighborhood for calculating video image.
The S102 the specific steps are：Neighborhood is chosen centered on the space-time characteristic point on the integral video of video image,
And by integrate video neighborhood subregion carry out small echo operation, obtain integral video all subregion small echo operation values.
It is to be appreciated that above-mentioned neighborhood at least has a sub-regions.
It illustrates herein and the subregion in neighborhood and field is introduced：Centered on space-time characteristic point, choosing size is
20σx×20σx×3σtRegion as neighborhood, neighborhood G ' is divided into 4 × 4 × 3 regions, which is subregion, divide
There are 5 σ in obtained every sub-regionsx×5σx×σtA pixel.Specifically, space-time characteristic neighborhood of a point G ' is represented by：
The small echo operation that step S102 is carried out is to utilize 2 σ of size to every sub-regionsx×2σx×σtHaar small echos
Template carries out response calculating, acquires x respectively, y, the response on the directions t.
S103, small echo operation values are rotated using the principal direction of space-time characteristic point, and is transported based on postrotational small echo
The response vector of all subregion of space-time characteristic point is calculated in calculation value.
To explain that second example in the present embodiment is incorporated herein in step S103：Assuming that integral video is in subregion
Small echo operation values include x, y, the small echo operation values F in the directions tx、FyAnd Ft, then in second example, step S103 is specially：
S1031, according to formula Fx'=- Fx×sin(θ)+Fy× cos (θ), Fy'=Fx×cos(θ)+Fy× sin (θ),
Ft'=Ft, to Fx、FyAnd FtIt is rotated, obtains integral video in subregion x, y, the postrotational small echo operation values F in the directions tx'、
Fy' and Ft'。
S1032, statistics integral video are in the postrotational small echo operation values F of subregion all directionsx'、Fy' and Ft', obtain ∑
Fx'、∑|Fx'|、∑Fy'、∑|Fx'|、∑Ft'、∑|Ft'|。
S1033, according to Vr=[∑ Fx',∑|Fx'|,∑Fy',∑|Fy'|,∑Ft',∑|Ft' |] space-time spy is calculated
The response vector V of sign point all subregionr, response vector VrIndicate the response vector of subregion r.
According to the above-mentioned second exemplary content, then the response vector V of the first subregion is obtained1For：
V1=[∑ F 'x,∑|F′x|,∑Fy,∑F′y|,∑F′t,∑F′t|]
Continue to illustrate with the example of above-mentioned illustration neighborhood：Since neighborhood shares 4 × 4 × 3 sub-regions, and it is every
One sub-regions generate 6 gradient statistical values, and (gradient statistical value herein is the ∑ F counted in step S1032x'、∑|
Fx'|、∑Fy'、∑|Fx'|、∑Ft'、∑|Ft' |), therefore proposed Feature Descriptor V shares 4 × 4 × 3 × 6=
288 D feature vectors form.In order to ensure space-time characteristic point description son contrast invariance, need to this 228 dimension when
Description of empty characteristic point is normalized, that is, carries out step S104.
S104, count space-time characteristic point all subregion response vector, and counted using the response vector after statistics
It calculates, obtains description of space-time characteristic point.
To explain that the third example in the present embodiment is incorporated herein in step S104, in the third example, step
S104 is specially：
The response vector for counting all subregion of space-time characteristic point, obtains the response vector V of space-time characteristic point, further
According toThe sub- V' of description of space-time characteristic point can be calculated, wherein | | V | | indicate the modulus value of response vector V.
In the present embodiment provides the description method of video image space-time characteristic point, description of space-time characteristic point is to be based on
Integral video obtains, and integrates half-tone information and movable information that video comprehensively contains video image, therefore the present invention
The description method of offer comprehensively uses video image information, and description method through the invention can be to avoid unilateral to regarding
The problem of frequency is characterized；Simultaneously as description method also rotates small echo operation values, so ensure that description
Rotational invariance is further realized the characterization to video image by description with rotational invariance, be can avoid unilateral
The problem of characterization to video；On the other hand, description of the invention method is to the region centered on space-time characteristic point
It carries out small echo operation and obtains small echo response, be based on small echo response later and be worth to description, by obtaining describing sub- institute's profit
Small echo response is not only the small echo response of space-time characteristic point, further includes in the region centered on space-time characteristic point
The small echo response of each pixel, therefore what obtained description can be more precisely to space-time characteristic point characterize and then to video
Image is characterized, to avoid the problem that unilateral being characterized to video.
Second embodiment
The present embodiment additionally provides a kind of image space-time characteristic point and describes device, shown in Figure 2 comprising processor
21, memory 22 and communication bus 23, wherein：
Communication bus 23 is for realizing the connection communication between processor 21 and memory 22；
Processor 21 is for executing the program stored in memory 22, to realize the video image in above-mentioned first embodiment
Each step of the description method of space-time characteristic point.
It should be noted that for each method embodiment above-mentioned, describe, therefore it is all expressed as a series of for simplicity
Combination of actions, but those skilled in the art should understand that, the present invention is not limited by the described action sequence because
According to the present invention, certain steps may be used other sequences or be carried out at the same time.Secondly, those skilled in the art should also know
It knows, embodiment described in this description belongs to preferred embodiment, and involved action and module might not all be this hairs
Necessary to bright.
In the above-described embodiments, it all emphasizes particularly on different fields to the description of each embodiment, there is no the portion being described in detail in some embodiment
Point, the associated description of other embodiments is may refer to, meanwhile, the embodiments of the present invention are for illustration only, do not represent
The quality of embodiment, those skilled in the art under the inspiration of the present invention, are not departing from present inventive concept and right is wanted
It asks under protected ambit, can also make many forms, all of these belong to the protection of the present invention.
Claims (10)
1. a kind of image characteristic point describes method, which is characterized in that the method includes：
Step A1, cylindrical region is constructed centered on the space-time characteristic point on the integral video of video image, and by the product
Divide video to carry out small echo operation in the predeterminable area of the cylindrical region, obtains the integral video in the cylindrical region
The small echo operation values of each predeterminable area, the small echo operation values of more each predeterminable area, by the maximum predeterminable area of small echo operation values
Corresponding direction is determined as the principal direction θ of the space-time characteristic point；The cylindrical region at least has there are one predeterminable area,
The integral video is obtained based on the half-tone information of each pixel and movable information in the video image；
Step A2, neighborhood is chosen centered on the space-time characteristic point on the integral video of video image, and by the integral video
Small echo operation is carried out in the subregion of the neighborhood, obtains the integral video in the small echo operation values of all subregion, the neighbour
Domain at least has a sub-regions；
Step A3, the small echo operation values of all subregion are rotated using the principal direction θ of the space-time characteristic point, and base
The response vector of all subregion of the space-time characteristic point is calculated in postrotational small echo operation values；
Step A4, the response vector of all subregion of the space-time characteristic point is counted, and is carried out using the response vector after statistics
It calculates, obtains description of the space-time characteristic point.
2. description method as described in claim 1, which is characterized in that small echo operation values packet of the integral video in subregion
Include Fx、FyAnd Ft, the Fx、Fy、FtThe integral video is indicated respectively in subregion x, y, the small echo operation values in the directions t, then institute
Stating step A3 includes：
According to formula Fx'=- Fx×sin(θ)+Fy× cos (θ), Fy'=Fx×cos(θ)+Fy× sin (θ), Ft'=Ft, to institute
State Fx、FyAnd FtIt is rotated, obtains the integral video in the subregion x, y, the postrotational small echo operation values in the directions t
Fx'、Fy' and Ft'；
The integral video is counted in the postrotational small echo operation values F of the subregion all directionsx'、Fy' and Ft', obtain ∑ Fx'、
∑|Fx'|、∑Fy'、∑|Fx'|、∑Ft'、∑|Ft'|；
According to Vr=[∑ Fx',∑|Fx'|,∑Fy',∑|Fy'|,∑Ft',∑|Ft' |] that the space-time characteristic point is calculated is each
The response vector V of subregionr, the response vector VrIndicate the response vector of subregion r.
3. description method as described in claim 1, which is characterized in that the step A4 includes：
The response vector for counting all subregion of the space-time characteristic point obtains the response vector V of the space-time characteristic point；
According toThe description sub- V' of the space-time characteristic point is calculated, described | | V | | indicate the mould of response vector V
Value.
4. description method as described in any one of claims 1-3, which is characterized in that the step A1 includes the following steps：
Step A11, cylindrical region is constructed centered on the space-time characteristic point on the integral video of video image, to preset size
Three-dimensional sector region slided according to default sliding type in the cylindrical region, and by the integral video in institute
State three-dimensional sector region and carry out small echo operation, obtain the integral video the three-dimensional sector region small echo operation values；
The three-dimensional sector region is slided according to the default sliding type, will obtain at least one in the cylindrical region
A three-dimensional sector region；
Step A12, each three-dimensional fan is calculated in the small echo operation values of each three-dimensional sector region using the integral video
The characteristic vector in shape region, the characteristic vector include modulus value and direction；
Step A13, the characteristic vector of each three-dimensional sector region described in comparison, by the maximum three-dimensional sector region characteristic vector of modulus value
Corresponding direction is determined as the principal direction θ of the space-time characteristic point.
5. description method as claimed in claim 4, which is characterized in that the three-dimensional sector region is the three-dimensional that subtended angle is π/3
Sector region；The default sliding type is to carry out rotational slide in the cylindrical region with 0.2 radian of step-length.
6. description method as claimed in claim 4, which is characterized in that the integral video is made of at least one pixel,
The integral video includes each picture of the integral video in the three-dimensional sector region in the small echo operation values of three-dimensional sector region
The small echo operation values of vegetarian refreshments；
The small echo operation values of the integral video image vegetarian refreshments include Ex、Ey, the Ex、EyThe pixel of the integral video is indicated respectively
Point is in x, the small echo operation values in the directions y；
Then the step A12 includes:
Using the small echo operation values of integral each pixel of video in the three-dimensional sector region, according toModulus value m is calculatedwWith direction θw, by (mw,θw) as institute
The characteristic vector of three-dimensional sector region is stated, the w indicates to calculate used three-dimensional sector region when characteristic vector.
7. a kind of image space-time characteristic point describes device, described device includes processor, memory and communication bus；
The communication bus is for realizing the connection communication between processor and memory；
The processor is for executing one or more program stored in memory, to realize following steps：
Step A1, cylindrical region is constructed centered on the space-time characteristic point on the integral video of video image, and by the product
Divide video to carry out small echo operation in the predeterminable area of the cylindrical region, obtains the integral video in the cylindrical region
The small echo operation values of each predeterminable area, the small echo operation values of more each predeterminable area, by the maximum predeterminable area of small echo operation values
Corresponding direction is determined as the principal direction θ of the space-time characteristic point；The cylindrical region at least has there are one predeterminable area,
The integral video is obtained based on the half-tone information of each pixel and movable information in the video image；
Step A2, neighborhood is chosen centered on the space-time characteristic point on the integral video of video image, and by the integral video
Small echo operation is carried out in the subregion of the neighborhood, obtains the integral video in the small echo operation values of all subregion, the neighbour
Domain at least has a sub-regions；
Step A3, the small echo operation values of all subregion are rotated using the principal direction θ of the space-time characteristic point, and base
The response vector of all subregion of the space-time characteristic point is calculated in postrotational small echo operation values；
Step A4, the response vector of all subregion of the space-time characteristic point is counted, and is carried out using the response vector after statistics
It calculates, obtains description of the space-time characteristic point.
8. extraction element as claimed in claim 7, the integral video includes F in the small echo operation values of subregionx、FyAnd Ft,
The Fx、Fy、FtIndicate the integral video in subregion x, y, the small echo operation values in the directions t respectively；
The processor is additionally operable to execute the program of the memory storage, to realize step A3, including：
According to formula Fx'=- Fx×sin(θ)+Fy× cos (θ), Fy'=Fx×cos(θ)+Fy× sin (θ), Ft'=Ft, to institute
State Fx、FyAnd FtIt is rotated, obtains the integral video in the subregion x, y, the postrotational small echo operation values in the directions t
Fx'、Fy' and Ft'；
The integral video is counted in the postrotational small echo operation values F of the subregion all directionsx'、Fy' and Ft', obtain ∑ Fx'、
∑|Fx'|、∑Fy'、∑|Fx'|、∑Ft'、∑|Ft'|；
According to Vr=[∑ Fx',∑|Fx'|,∑Fy',∑|Fy'|,∑Ft',∑|Ft' |] that the space-time characteristic point is calculated is each
The response vector V of subregionr, the response vector VrIndicate the response vector of subregion r.
9. extraction element as claimed in claim 7 or 8, which is characterized in that the processor is additionally operable to execute the memory
The program of storage, to realize step A1, including：
Step A11, cylindrical region is constructed centered on the space-time characteristic point on the integral video of video image, to preset size
Three-dimensional sector region slided according to default sliding type in the cylindrical region, and by the integral video in institute
State three-dimensional sector region and carry out small echo operation, obtain the integral video the three-dimensional sector region small echo operation values；
The three-dimensional sector region is slided according to the default sliding type, will obtain at least one in the cylindrical region
A three-dimensional sector region；
Step A12, each three-dimensional fan is calculated in the small echo operation values of each three-dimensional sector region using the integral video
The characteristic vector in shape region, the characteristic vector include modulus value and direction；
Step A13, the characteristic vector of each three-dimensional sector region described in comparison, by the maximum three-dimensional sector region characteristic vector of modulus value
Corresponding direction is determined as the principal direction θ of the space-time characteristic point.
10. extraction element as claimed in claim 9, which is characterized in that the integral video is made of at least one pixel,
The integral video includes each picture of the integral video in the three-dimensional sector region in the small echo operation values of three-dimensional sector region
The small echo operation values of vegetarian refreshments；
The small echo operation values of the integral video image vegetarian refreshments include Ex、Ey, the Ex、EyThe pixel of the integral video is indicated respectively
Point is in x, the small echo operation values in the directions y；
Then the processor is additionally operable to execute the program of the memory storage, to realize step A12, including：
Using the small echo operation values of integral each pixel of video in the three-dimensional sector region, according toModulus value m is calculatedwWith direction θw, by (mw,θw) as institute
The characteristic vector of three-dimensional sector region is stated, the w indicates to calculate used three-dimensional sector region when characteristic vector.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810488159.0A CN108764254B (en) | 2018-05-21 | 2018-05-21 | Image feature point description method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810488159.0A CN108764254B (en) | 2018-05-21 | 2018-05-21 | Image feature point description method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108764254A true CN108764254A (en) | 2018-11-06 |
| CN108764254B CN108764254B (en) | 2021-10-22 |
Family
ID=64008540
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810488159.0A Active CN108764254B (en) | 2018-05-21 | 2018-05-21 | Image feature point description method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108764254B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120008868A1 (en) * | 2010-07-08 | 2012-01-12 | Compusensor Technology Corp. | Video Image Event Attention and Analysis System and Method |
| CN102750724A (en) * | 2012-04-13 | 2012-10-24 | 广州市赛百威电脑有限公司 | Three-dimensional and panoramic system automatic-generation method based on images |
| CN103218825A (en) * | 2013-03-15 | 2013-07-24 | 华中科技大学 | Quick detection method of spatio-temporal interest points with invariable scale |
| CN107918927A (en) * | 2017-11-30 | 2018-04-17 | 武汉理工大学 | A kind of matching strategy fusion and the fast image splicing method of low error |
-
2018
- 2018-05-21 CN CN201810488159.0A patent/CN108764254B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120008868A1 (en) * | 2010-07-08 | 2012-01-12 | Compusensor Technology Corp. | Video Image Event Attention and Analysis System and Method |
| CN102750724A (en) * | 2012-04-13 | 2012-10-24 | 广州市赛百威电脑有限公司 | Three-dimensional and panoramic system automatic-generation method based on images |
| CN103218825A (en) * | 2013-03-15 | 2013-07-24 | 华中科技大学 | Quick detection method of spatio-temporal interest points with invariable scale |
| CN107918927A (en) * | 2017-11-30 | 2018-04-17 | 武汉理工大学 | A kind of matching strategy fusion and the fast image splicing method of low error |
Non-Patent Citations (3)
| Title |
|---|
| HUA-BIN WANG: "Novel Algorithm for Hand Vein Feature Extraction and Recognition Based on Vertical Integral Projection and Wavelet Decomposition", 《2012 2ND INTERNATIONAL CONFERENCE ON CONSUMER ELECTRONICS, COMMUNICATIONS AND NETWORKS (CECNET)》 * |
| 李岩山: "基于局部不变特征的交通异常视频图像检测的关键技术研究", 《中国博士学位论文全文数据库信息科技辑》 * |
| 李彬: "全景视频实时拼接技术研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108764254B (en) | 2021-10-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108961303B (en) | Image processing method and device, electronic equipment and computer readable medium | |
| US9429418B2 (en) | Information processing method and information processing apparatus | |
| US10455152B2 (en) | Panoramic video processing method and device and non-transitory computer-readable medium | |
| EP2903256B1 (en) | Image processing device, image processing method and program | |
| US20180247132A1 (en) | System and method for person counting in image data | |
| CN110136156A (en) | A kind of polygonal region detection method and device | |
| Shen et al. | Real-time and robust compressive background subtraction for embedded camera networks | |
| CN104657709B (en) | Facial image recognition method, device and server | |
| WO2019237745A1 (en) | Facial image processing method and apparatus, electronic device and computer readable storage medium | |
| JP7383714B2 (en) | Image processing method and device for animal faces | |
| CN110728644B (en) | Image generation method and device, electronic equipment and readable storage medium | |
| CN113538587A (en) | Camera coordinate transformation method, terminal and storage medium | |
| CN110751602B (en) | Conformal distortion correction method and device based on face detection | |
| JP2020525959A (en) | Method, device, medium and equipment for determining the bounding box of an object | |
| CN111357034A (en) | Point cloud generation method, system and computer storage medium | |
| CN106503682A (en) | Crucial independent positioning method and device in video data | |
| CN104765440B (en) | Hand detection method and equipment | |
| CN113793259B (en) | Image zooming method, computer device and storage medium | |
| JP5704909B2 (en) | Attention area detection method, attention area detection apparatus, and program | |
| CN106997580B (en) | Picture processing method and device | |
| CN113610835A (en) | Human shape detection method for nursing camera | |
| CN108764254A (en) | A kind of image characteristic point describes method | |
| WO2021109863A1 (en) | Photo processing method and photo processing apparatus | |
| CN102843479A (en) | File scanning method, file scanning device and portable electronic device | |
| JP2018092507A (en) | Image processing apparatus, image processing method, and program |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for 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: 20230423 Address after: 518000 26-1, block C, World Trade Plaza, Fuhong Road, Futian street, Shenzhen City, Guangdong Province Patentee after: SHENZHEN XINGHAI IOT TECHNOLOGY Co.,Ltd. Address before: 518060 No. 3688 Nanhai Road, Shenzhen, Guangdong, Nanshan District Patentee before: SHENZHEN University |