Texture Modeling of Image Data

Abstract

Texture modeling techniques for image data are described. In one or more implementations, texels in image data are discovered by one or more computing devices, each texel representing an element that repeats to form a texture pattern in the image data. Regularity of the texels in the image data is modeled by the one or more computing devices to define translations and at least one other transformation of texels in relation to each other.

What is claimed is: 
     
         1 . A method comprising:
 discovering texels in image data by one or more computing devices, each of the texels representing an element that repeats to form a texture pattern in the image data; and   modeling regularity of the texels in the image data by the one or more computing devices to define translations and at least one other transformation of texels in relation to each other.   
     
     
         2 . A method as described in  claim 1 , wherein the discovering is performed by finding self-similar patterns. 
     
     
         3 . A method as described in  claim 1 , wherein the discovering is performed by using a patch matching technique. 
     
     
         4 . A method as described in  claim 3 , wherein the patch matching technique includes finding a nearest neighbor field (NNF) for patches in the image data. 
     
     
         5 . A method as described in  claim 1 , wherein the at least one other transformation includes scaling, skewing, rotation, or perspective distortion. 
     
     
         6 . A method as described in  claim 1 , wherein the modeling includes estimating a homography between planes that include respective ones of the texels. 
     
     
         7 . A method as described in  claim 1 , wherein the modeling includes fitting a lattice based on the discovered texels to the image data. 
     
     
         8 . A method as described in  claim 7 , wherein the fitting of the lattice includes representing translational symmetric patterns as a pair of translation vectors and voting to determine which of the translation vectors are to be used. 
     
     
         9 . A method as described in  claim 8 , wherein the voting is based at least in part on a count of inlier points that are within a threshold of an integer position such that the pair of translation vectors having a corresponding highest inlier count in relation to other pairs of translation vectors is used as basis vectors for the fitting of the lattice. 
     
     
         10 . A method as described in  claim 1 , further comprising performing one or more texture manipulation editing operations at a texel level based on the generated model. 
     
     
         11 . A method comprising:
 utilizing a patch matching technique by one or more computing devices to locate self-similar patterns involving patches including scales, rotations, skews, or perspective distortions within image data; and   generating a model by the one or more computing devices by fitting a lattice to the located self-similar patterns.   
     
     
         12 . A method as described in  claim 11 , wherein the patch matching technique is used to locate texels that describe an element that repeats to form the texture pattern in the image data. 
     
     
         13 . A method as described in  claim 11 , wherein the patch matching technique includes an initialization stage at which patches are given random assignment, a nearest neighbor stage in which involves propagation of matches, and a search stage that involves a random search for improvement in concentric neighborhoods. 
     
     
         14 . A system comprising:
 a texture pattern discovery module implemented at least partially in hardware, the texture pattern discovery module configured to discover texels in image data, each of the texels representing an element that repeats to form a texture pattern in the image data; and   a model generation module implemented at least partially in hardware, the model generation module configured to generate a model describing regularity of the texels in the image data to define translations and at least one other transformation of texels in relation to each other.   
     
     
         15 . A system as described in  claim 14 , wherein the texture pattern discovery module is configured to perform the discovering by finding self-similar patterns. 
     
     
         16 . A system as described in  claim 14 , wherein the texture pattern discovery module is configured to perform the discovering using a patch matching technique. 
     
     
         17 . A system as described in  claim 16 , wherein the patch matching technique includes finding a nearest neighbor field (NNF) for patches in the image data. 
     
     
         18 . A system as described in  claim 14 , wherein the at least one other transformation includes scaling, skewing, or rotation. 
     
     
         19 . A system as described in  claim 14 , wherein the model generation module is configured to generate the model by estimating a homography between planes that include respective ones of the texels. 
     
     
         20 . A system as described in  claim 14 , wherein the model generation module is configured to generate the model by fitting a lattice based on the discovered texels to the image data.