Quad tree based Image compression
Shekhar Verma, Amit Kant Pandit · 2008
In this paper, variable image segmentation and mixed transform are used to capitalize on the narrow band and broadband signal components for image compression. A mixture of transforms can exploit the different spectral bands in an image in achieve higher compression as compared to single transforms. However, the optimal distribution of the signal into transforms. However, the optimal distribution of the signal into transform specific spectral sub bands is computationally intensive and requiring multiple switching between time and transform domains. The present work tries to achieve this segregation by the decomposing the image into variable size blocks using a quad tree. The quad tree is formed with 4times4, 8times8 and 16times16 blocks. It is pruned by fusing smaller blocks according to the smoothness in different region of the image. The blocks in the pruned quad tree are preprocessed by removing the DC coefficient of DCT from each pixel of the block followed by DCT-WHT transform to capitalize on the residual image. Simulation results show that the image tree is segmented has bigger blocks for smooth regions with smaller blocks in regions with large variation content. Moreover, for similar compression rations, a gain of around 1.7-5 dB in PSNR over conventional JPEG was observed.