Today's lossless image compression algorithms are not well optimized for high frequency components. These occure especially in textured regions of and image and make image compression ineffective. On the other hand the high demand of storage is not justified since the human visual system perceives texture with regard to it's general structure while omitting specific details. Rather than storing textured parts of an image we present a method for analyzing (in the source image) and resynthesizing (in the decompressed target image) textured regions with regard to it's structural appearance. First we define image blocks in a way that they cover regions with homogenious texture in the source image. These blocks are transformed each in turn. For the transform we use the Karhunen-Loève transform. After the transform into the new domain we statistically analyze the resulting coefficients. For resynthesizing new texture we generate random numbers that exactly meet these statistics. Using the inverse KL-transform the random coefficients are finally transformed back into the spatial domain. The visual appearance of the resulting artificial texture matches the original to a very high degree.
Dieser Eintrag ist Teil der Universitätsbibliographie.
Search Authors in