| This thesis is concerned with a shape adaptive wavelet image coding scheme based on Human Visual System (HVS) model to code arbitrarily shaped objects or regions in still colour images or video sequences. The main difference between the proposed scheme and JPEG 2000 is that the former codes the given image through object by object while the later through tiling. A wavelet transform for directly coding the arbitrarily shape objects is implemented. A contrast sensitivity-based embedded zerotree wavelet (CS-EZW) is developed. Through experimental testing, compared with the other coders, including JPEG, shape adaptive DWT, this proposed coding system gives better results in terms of visual quality of reconstructed images. The performance of this system is evaluated by using three evaluation criteria, including peak signal-to-noise ratio (PSNR), viewing by subjects and Frequency Shell Correlation (FSC). |
