EE Seminar: Co-occurrence Based Texture Synthesis
Speaker: Anna Darzi
M.Sc. student under the supervision of Prof. Shai Avidan
Wednesday, March 6th, 2019 at 15:30
Room 011, Kitot Bldg., Faculty of Engineering
Co-occurrence Based Texture Synthesis
Abstract
We model local texture patterns using the co-occurrence statistics of pixel values. We then train a conditional generative adversarial network (cGAN) to synthesize new textures from the co-occurrence statistics and a random seed noise. Co-occurrences have long been used to measure similarity between textures. That is, two textures are considered similar if their corresponding co-occurrence matrices are similar. By the same token, we show that multiple textures generated from the same co-occurrence matrix are similar to each other. This gives rise to a new texture synthesis algorithm.
We use co-occurrence based texture synthesis in various settings. For example, we generate variations on the input texture by using the same co-occurrence statistics with different seed noise, or we merge two co-occurrence matrices to smoothly
interpolate between different textures.
In another case, we synthesize a dynamic texture sequence by interpolating between two co-occurrence matrices. Yet another option is to create a sequence that
summarizes the various local texture patterns in a given texture image. And because co-occurrence statistics have clear and intuitive meaning we develop a tool that lets users modify them directly and hence influence the local characteristics of the synthesized texture image.
