Embedding Secret Data in Digital Media Using Texture Synthesis

Literature Survey

A simple and efficient pixel based method is presented by Efros and Leung (Efros & Leung, 1999) in 1999. This method generates the synthesized image pixel by pixel and use spatial neighborhood comparisons to choose the most similar pixel in a sample texture as the output pixel. In this, a single pixel is generated at a time from an initial seed. A fixed size window with user specified size is taken which is centered on the currently synthesizing pixel. The more matching pixel is searched and copied in the output target. This process is repeated until all pixels in the target image are generated. But any wrongly synthesized pixels during the process influence the rest of the result causing propagation of errors.

Wei and Levoy (2000) presented a deterministic algorithm and the output texture is generated in a scan line order. This method improves the speed of the synthesis procedure by using tree structured vector quantization (TSVQ) to match the target pixel with the neighborhood pixels. This fastest algorithm has a largest memory requirement. But it failed to synthesize the texture images of flowers, leafs, pebbles etc.

L. Liang, C. Liu(2001) presented an algorithm for synthesizing textures from an input sample. This patch-based sampling algorithm is very fast and it creates high-quality texture image. This algorithm works well for a wide variety textures like regular to stochastic textures. Can be sampling patches using a nonparametric estimation of the local conditional MRF density function. Also avoid mismatching features across patch boundaries of an image. The building blocks of the patch-based sampling algorithm are patches of the input sample texture to construct the synthesized texture. We can carefully select these patches of the input sample texture and paste it into the synthesized texture to avoid mismatching features across patch boundaries. Patch-based sampling algorithm combines the nonparametric sampling and patch pasting strengths. The texture patches in the sampling scheme provide implicit constraints to avoid garbage found in some textures.