Compression Techniques essays

Compression Techniques essays This research describes an image representation technique that entails progressive refinement of user specified regions of interest (ROI) of large images. Progressive refinement of original quality can be accomplished in theory. However, due to heavy burden on storage resources for our applications, we restrict the refinement to about 25% of the original data resolution. Wavelet decomposition with Vector Quantization (VQ) of the high frequency components and JPEG/DCT compression of low frequency component is used as representation framework. Our software will reconstruct the region selected by the user from its wavelet decomposition at desired resolution. Further refinement from the first preview can be obtained progressively by transmitting high frequency coefficients from low resolution to high resolution, which are compressed by variant of Vector Quantization called Model Based Vector Quantization. The user will have an option of progressive build up of the ROIs until full resolution stored or terminate the transmission at any time during the progressive refinement. The entire architecture of the program is based on object oriented programming using C++. A multiresolution decomposition into wavelet coefficients provides the most useful image representation or image browsing. Image decomposition is done recursively into wavelet coefficients using the technique [2] shown in Figure 1. The given image is decomposed into low and high frequency bands along the rows and columns. The high frequency subbands are first scalar quantized and the low frequency band, LL subband of the highest-level decomposition is compressed using JPEG/DCT technique. The scalar quantization is heavier for higher resolutions and reduced successively for low resolution high frequency components. The coefficients are then further compressed by a variant of Vector ...