US2013077870A1PendingUtilityA1

Evaluation of edge direction information

Assignee: FRAUNHOFER GES FORSCHUNGPriority: Sep 4, 2000Filed: Nov 26, 2012Published: Mar 28, 2013
Est. expirySep 4, 2020(expired)· nominal 20-yr term from priority
G06V 40/161G06T 7/75G06V 40/10G06K 9/4638
40
PatentIndex Score
0
Cited by
0
References
0
Claims

Abstract

A physically demarcated body part is recognized and located using only a relatively small amount of computation, but with a sufficient degree of recognition accuracy. For this purpose a procedure is proposed for detecting physically demarcated body parts (face, hand, leg) of a person's image ( 5 ) if a body part ( 2 ) as depicted in front of a background ( 3 ). Borderlines ( 5 d, 5 e ) in the image are only evaluated along line directions ( 5 a′, 4 a′, 4 b′, 5 c ′) to determine, by comparing with model ( 30 ), whether the body part image corresponds to a type of body part given by the model. In addition, line directions ( 5 d′, 5 e ′) inside a body part image and borderline directions ( 5 a ) of a physically demarcated body part are used to locate and store its position.

Claims

exact text as granted — not AI-modified
What is claimed is: 
     
         1 - 27 . (canceled) 
     
     
         28 . A method for recognizing a physically delimited body type portion of a person pictorially represented in a first image wherein at least one not yet recognized a body type portion is contained in the first image in front of a background, a position of the body type portion is to be localized by at least positional information thereof to be obtained and stored, the method comprising the steps of:
 (a) converting the first image through at least one intermediate image into an edge image having edge lines accentuated and evaluated in a following step providing an edge direction image, containing directions of said edge lines;   (b) providing an edge line model containing other edge line information as to directions, the other edge line direction information corresponding to the body type portion to be recognized   (c) wherein only edge line directions in said edge direction image are evaluated in comparison with the edge line model containing the other edge line information, to determine whether the body type portion appearing in front of the background corresponds to the body type portion to be recognized;   (d) wherein positional information is determined in said first image and said edge direction image provided that the type of body portion in said edge direction image corresponds to the yet unidentified body type portion.   
     
     
         29 . The method according to  claim 28 , wherein images of several same body type portions are present on said first image, and are consecutively compared with the edge line model containing the other edge line direction information according to edge line information thereof. 
     
     
         30 . The method according to  claim 28 , wherein the positional information contains data about a position of the yet unidentified body type portion in the edge direction image and correspondingly in said first image. 
     
     
         31 . The method according to  claim 28 , wherein both first edge lines inside the pictorially represented body type portion, as well as an edge line of the physically delimited body type portion with directions are provided. 
     
     
         32 . The method according to  claim 28 , wherein the first image comprises a scenic picture having different, but locally unchanging intensities and appearing as a static image in a digital form. 
     
     
         33 . The method according to  claim 28  wherein the first image comprises a scenic image having locally changing intensities and shows a subject ( 1 ) moving in front of the background ( 3 ) such that one of a relative position and an orientation of at least one depicted body part changes with time. 
     
     
         34 . The method according to  claim 28  wherein a boundary line in the first image comprises edge information that is associated with orientation. 
     
     
         35 . The method according to  claim 28  comprising accentuating borderlines from an evaluation of intensity difference between two adjacent image areas to determine contrast. 
     
     
         36 . The method according to  claim 28  wherein the body type portion is represented by a model ( 30 ) and is represented in front of a background ( 3 ) in an image ( 5 ) and is made available for binary comparison for determining correspondence between the first image and the edge line model. 
     
     
         37 . The method according to  claim 28  wherein the first image comprises a grayscale image ( 5 ) having at least one component, in one of a visible light range and an invisible wavelength range. 
     
     
         38 . The method according to  claim 28  further comprising determining and storing size information in addition to the positional information of the body type portion. 
     
     
         39 . The method according to  claim 38  wherein the size information is determined by changing a relative size of the edge line model and the first image ( 5 ) and comparing the model and the first image as thereby sized, for determining size information. 
     
     
         40 . The method according to  claim 28  wherein the intermediate image has been evaluated in terms of direction compared with the edge line model by constructing sums of angle oriented similarity values. 
     
     
         41 . The method according to  claim 40  wherein a threshold value is made available which is compared to all similarity values distributed over the first image, only concluding that a sought-for body type part is present in the first image if at least one index is below a minimum threshold value. 
     
     
         42 . A method for detection of a physically demarcated body part of a person's image ( 5 ) in which a body part ( 2 ) is depicted in front of a background ( 3 ), wherein:
 borderlines of the image are evaluated only according to directions to determine by comparison with an edge line model ( 30 ) whether the body part in the image corresponds to a type represented by the edge line model, wherein   at least borderline directions inside at least one body part image and borderline directions ( 5 A) of at least one physically demarcated body part in the image are used for locating a position and wherein corresponding positional information is stored.   
     
     
         43 . A preparation method for detection of a physically demarcated body part in at least one image wherein a body part ( 2 ) is depicted in front of a background and directional information for at least one of the body part and a background thereof are prepared by calculation, the directional information being sufficient to compare whether the body part in the image corresponds to a type given by an edge line model, wherein:
 (a) the at least one body part image is converted into a further representation comprising directional data for borderlines between the body part and the background, wherein the directional data has a clarity value for each direction;   (b) the further representation is digitally processed to produce a working image by using a threshold value to suppress directional data at positions where the clarity is below the threshold value, resulting in reduced directional information in the working image.   
     
     
         44 . The method according to  claim 43  wherein remaining directional information stays unchanged in the working image, but the clarity values are set at a standard value for all remaining directions to neutralize clarity information in the working image. 
     
     
         45 . The method according to  claim 44  wherein the remaining directional information and remaining directions are evaluated in a next step by using an edge line model ( 30 ) and wherein a position of the body part is located and positional information for said body part is stored. 
     
     
         46 . The method according to  claim 43  wherein a plane of a representation which has been processed with threshold values, is processed with a binary operator to obtain a pattern image which is combined with the further representation. 
     
     
         47 . The method according to  claim 40  wherein only a reduced number of terms are used to obtain the sums of similarity values for each relative position of the edge line model. 
     
     
         48 . The method according to  claim 40  wherein angles for the directional information between 90° and 180° are converted to corresponding values between 90° and 0° before being compared, to compare only acute angles. 
     
     
         49 . The method according to  claim 47  wherein angles for the directional information between 90° and 180° are converted to corresponding values between 90° and 0° before being compared, to compare only acute angles. 
     
     
         50 . The method according to  claim 48  wherein the edge line model contains only directional angles between 0° and 90° for borderlines in the edge line model. 
     
     
         51 . The method according to  claim 49  wherein the edge line model contains only directional angles between 0° and 90° for borderlines in the edge line model.

Join the waitlist — get patent alerts

Track US2013077870A1 — get alerts on status changes and closely related new filings.

We store only your email — no account needed. See our privacy policy.