| Some of the most common still images encountered on computer displays are palette color images which have a limited number of color components. Examples of such spectrally limited color images are color-enhanced weather maps, clipart (illustrations), business graphics (bar, line, pie charts), and color quantized images (icons). Although these spectrally limited images are common, most of the literature focuses on continuous-tone images. Compression algorithms designed for continuous-tone images (both lossy and lossless) perform poorly on these spectrally limited images. In this dissertation, multiple-pass lossless context-based compression algorithms that utilize the spatial correlation of the color components were investigated.; The main contributions are, (i) a two-part context coding algorithm that avoids the zero-frequency problem, (ii) an equivalent adaptive coding algorithm that also mitigates the zero-frequency problem, (iii) a context model statistics coder that exploits the statistics' interdependencies, (iv) a context-based data sequence coder that exploits the dependencies between context states and prior coded symbols, (v) a compact representation for the statistics of large contexts, and (vi) a very compact searchable representation of static binary trees. |
