Abstract
Tzung-ruen YehMany lossy image compression algorithms have been developed for low bit rate applications. Among them, quadtree decomposition is a simple technique for image representation at different resolution levels that adapts spatially, allocating more bits to those complicated areas in the image. Its simplicity and potential for relatively good performance for low bit rates makes quadtree a desirable choice for a compression scheme. A new scheme for quadtree code is proposed in this thesis. This scheme combines a new developed quadtree code with regular quadtree code. This new quadtree code is complemental to the regular one. Its usage and limitation is also discussed. The algorithms and experimental results are given. Experimental results show that the new algorithm requires about 1.8 seconds to encode a 512x512 gray level image and requires about 0.3 seconds to decode it while maintains a satisfactory image fidelity.