Texture synthesis has been an active and important research area in computer graphics and image processing. The goal of texture synthesis is to generate a new texture image from a texture sample such that the new mimic texture image is sufficiently different and appears as the same underlying stochastic process as the original texture sample....[ Read more ]

Texture synthesis has been an active and important research area in computer graphics and image processing. The goal of texture synthesis is to generate a new texture image from a texture sample such that the new mimic texture image is sufficiently different and appears as the same underlying stochastic process as the original texture sample.

Traditional texture synthesis techniques require prohibitively large amount of computer storage and synthesis time to generate a limited size of texture image. This thesis addresses an efficient and simple enough technique for synthesizing textures that facilitates memory efficient textures with an infinite large of size. The synthesizing process generates an initial texture block from the texture sample and subsequently all other texture blocks by measuring the joint occurrence of texture discrimination features. A final texture image is then synthesized by sampling consecutive spatial frequency bands from the texture sample.

Based on the assumption of spatial locality, texture images synthesized with this method are visually indistinguishable from the texture samples while preserving local patterns for a wide variety of deterministic and stochastic textures.