Transform coding of compound images using matching pursuit

Mixed Raster Content (MRC) coding is an efficient way of coding compound images. The layered model used gives rise to missing data in the foreground and background layers. When using a block-based transform for coding, the usual solution has been to fill in the missing data using some form of interpolation. In this paper we instead present a method using matching pursuit to find the transform coefficients. The presented method gives a gain of up to 1 dB on the tested images, compared to common data filling methods.