Detection and recognition of objects lacking textures

Abstract

Various embodiments provide methods and systems for detecting one or more segments of an image that are related to a particular object in the image (e.g., a logo or trademark) and extracting at least one feature point, each of which is represented by one feature point descriptor, based at least upon a contour curvature of the one or more segments. The at least one feature point descriptor can be converted into one or more codewords to generate a codeword database. A discriminative codebook can then be generated based upon the codeword database and utilized to detect objects and/or features in a query image.

1 . (canceled) 
     
     
         2 . A computer-implemented method, comprising:
 detecting, in image data, a shape representative of a brand or source;   determining a set of feature points of the shape;   generating respective feature point descriptors for individual feature points of the set of feature points;   converting the respective feature point descriptors into one or more codewords; and   generating a discriminative codebook based at least upon the one or more codewords.   
     
     
         3 . The computer-implemented method of  claim 2 , wherein the set of feature points are extracted from a contour curvature of the segment by using at least one of a plurality of feature detection techniques including maximally stable extremal regions (MSER), image moment, scale-invariant feature transform (SIFT), support vector machine (SVM), Adaboost, and Pyramid match kernel (PMK), Hessian-Affine, Harris-Affine, edge-based region (EBR), and scale-invariant shape features (SISF). 
     
     
         4 . The computer-implemented method of  claim 2 , further comprising:
 identifying a plurality of segments of the image data;   combining the plurality of segments into a cluster region; and   extracting the set of feature points based upon a contour curvature of the cluster region.   
     
     
         5 . The computer-implemented method of  claim 4 , wherein the cluster region includes two or more segments being within a predetermined distance with each other. 
     
     
         6 . The computer-implemented method of  claim 5 , further comprising:
 normalizing the predetermined distance based upon a size of the image.   
     
     
         7 . The computer-implemented method of  claim 4 , further comprising:
 determining a center of the contour curvature by averaging locations of all points on the contour curvature;   determining a signature for each point on the contour curvature by calculating a distance between the corresponding point to the center of the contour curvature;   determining one or more extremes, each of the extremes being a signature of points on the contour curvature; and   extracting the set of feature points according to the one or more extremes.   
     
     
         8 . The computer-implemented method of  claim 7 , further comprising:
 determining an orientation for each of the respective feature point descriptors based upon a direction from a corresponding feature point to the center of contour curvature.   
     
     
         9 . The computer-implemented method of  claim 2 , wherein the respective feature point descriptors are represented by at least one Log-polar shaping context descriptor. 
     
     
         10 . The computer-implemented method of  claim 2 , wherein the respective feature point descriptors are represented by at least one gray-scale window, a vector of filter outputs, or brightness at a single pixel. 
     
     
         11 . The computer-implemented method of  claim 2 , wherein the one or more codewords represent a particular appearance of the shape, and the one or more codewords are stored in a database that includes at least one codeword converted from corresponding feature point descriptors that represent another appearance of the shape, the another appearance being different from the particular appearance. 
     
     
         12 . The computer-implemented method of  claim 11 , further comprising:
 generating a set of different appearances of the shape by using random perspective transformations;   computing the at least one codeword based at least upon the set of feature points, the corresponding feature point descriptors, or coordinates of the set of feature points; and   generating the discriminative codebook based at least upon information stored in the database.   
     
     
         13 . The computer-implemented method of  claim 2 , further comprising:
 assigning a particular weight to each codeword stored in the database based upon uniqueness of the corresponding codeword comparing with other codewords in the database.   
     
     
         14 . A system, comprising:
 at least one processor; and   memory including instructions that, when executed by the at least one processor, cause the system to:
 detect, in image data, a shape representative of a brand or source; 
 determine a set of feature points of the shape; 
 generate respective feature point descriptors for individual feature points of the set of feature points; 
 convert the respective feature point descriptors into one or more codewords; and 
 generate a discriminative codebook based at least upon the one or more codewords. 
   
     
     
         15 . The system of  claim 14 , wherein the instructions when executed further cause the system to:
 identify a plurality of segments in the image data;   combine the plurality of segments into a cluster region; and   extract the set of feature points based upon a contour curvature of the cluster region.   
     
     
         16 . The system of  claim 15 , wherein the instructions when executed further cause the system to:
 determine a center of the contour curvature by averaging locations of all points on the contour curvature;   determine a signature for each point on the contour curvature by calculating a distance between the corresponding point to the center of the contour curvature;   determine one or more extremes, each of the extremes being a signature of points on the contour curvature; and   extract the at least one feature point according to the one or more extremes.   
     
     
         17 . The system of  claim 14 , wherein the instructions when executed further cause the system to:
 generate a set of different appearances of the shape by using random perspective transformations;   compute the at least one codeword based at least upon the set of feature points, the corresponding feature point descriptors, or coordinates of the set of feature points; and   generate the discriminative codebook based at least upon information stored in the database.   
     
     
         18 . A non-transitory computer-readable storage medium including instructions that, when executed by at least one processor of a computing system, cause the computing system to:
 detect, in image data, a shape representative of a brand or source;   determine a set of feature points of the shape;   generate respective feature point descriptors for individual feature points of the set of feature points;   convert the respective feature point descriptors into one or more codewords; and   generate a discriminative codebook based at least upon the one or more codewords.   
     
     
         19 . The non-transitory computer-readable storage medium of  claim 18 , wherein the instructions when executed further cause the system to:
 identify a plurality of segments in the image data;   combine the plurality of segments into a cluster region; and   extract the plurality of feature points based at least upon a contour curvature of the cluster region.   
     
     
         20 . The non-transitory computer-readable storage medium of  claim 18 , wherein the instructions when executed further cause the system to:
 determine a center of the contour curvature by averaging locations of all points on the contour curvature;   determine a signature for each point on the contour curvature by calculating a distance between the corresponding point to the center of the contour curvature;   determine one or more extremes, each of the extremes being a signature of points on the contour curvature; and   extract the plurality of feature points according to the one or more extremes.   
     
     
         21 . The non-transitory computer-readable storage medium of  claim 18 , wherein the instructions when executed further cause the system to:
 generate a set of different appearances of the shape by using random perspective transformations;   compute the at least one codeword based at least upon the plurality of feature points, the corresponding feature point descriptors, or coordinates of the plurality of feature points; and   generate the discriminative codebook based at least upon information stored in the database.