Systems and Methods for Providing Non-Parametric Texture Synthesis of Arbitrary Shape and/or Material Data in a Unified Framework

A non-transitory machine readable medium containing processor instructions, where execution of the instructions by a processor causes the processor to perform a process comprising:

• receiving data describing the exemplar object, the data including a set of source graphic objects that each contain a pixel at an index that contributes to an overall appearance vector at a location associated with the index, wherein the data includes a plurality of appearance vectors that define a desired texture;
• generating a feature vector for each pixel in a plurality of pixels of the exemplar object, the generating of a feature vector for a pixel including combining a set of appearance vectors surrounding the pixel and transforming the set through a convolutional neural network;
• determining a list of nearest neighbor candidates for each particular pixel in the plurality of pixels of the exemplar object using the feature vectors from the exemplar object, wherein each nearest neighbor candidate matches the particular pixel within a matching metric;
• and
• synthesizing, using at least the determined list of nearest neighbor candidates, an output graphic object by modifying each pixel in a plurality of pixels of the output graphic object to have an appearance vector which is within a similarity metric of an appearance vector of the particular pixel from the exemplar object.

A method for performing non-parametric texture synthesis of texture or material data taken from an exemplar object, the method comprising:

• receiving data describing the exemplar object, the data including a set of source graphic objects that each contain a pixel at an index that contributes to an overall appearance vector at a location associated with the index, wherein the data includes a plurality of appearance vectors that define a desired texture;
• generating a feature vector for each pixel in a plurality of pixels of the exemplar object, the generating of a feature vector for a pixel including combining a set of appearance vectors surrounding the pixel and transforming the set through a convolutional neural network;
• determining a list of nearest neighbor candidates for each particular pixel in the plurality of pixels of the exemplar object using the feature vectors from the exemplar object, wherein each nearest neighbor candidate matches the particular pixel within a matching metric;
• and
• synthesizing, using at least the determined list of nearest neighbor candidates, an output graphic object by modifying each pixel in a plurality of pixels of the output graphic object to have an appearance vector which is within a similarity metric of an appearance vector of the particular pixel from the exemplar object.