Quadtree-based Error Diffusion Block Truncation Coding

• Main Authors: Chih-LingHsu, 徐芷翎
• Other Authors: Pei-Yin Chen
• Format: Others
• Language: en_US
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/95173957923096646540

碩士 === 國立成功大學 === 資訊工程學系碩博士班 === 101 === In the age of information explosion, reducing the storage and bandwidth needed to store and transmit the images efficiently has become one important issue. Hence, methods to compress the image data are essential nowadays. Block Truncation Coding (BTC) is a simple and efficient technique. However the bit rate of the original BTC algorithm is relatively high compared to modern compression techniques such as JPEG or JPEG2000. Some investigations have been proposed to further reduce the bit rate so far. Nevertheless, these algorithms produce some annoying image artifacts caused by the low bit rate configuration and some even sacrifice the low-complexity characteristic of BTC. Hence, a low-complexity BTC technique that can reduce the perceptual artifacts effectively and provide good image quality at the low bit rate is crucial. Two image compression algorithms are proposed on the basis of halftoning-based BTC in this thesis. We apply the concept of quadtree decomposition to the proposed methods. The non-overlapping blocks of an image are segmented into smaller blocks based on the texture of the image. To reduce the blocking effect, we utilize the error diffusion technique which diffuses the quantized error into neighboring unprocessed pixels to maintain the local gray level. An adaptive scheme which is used to decide the thresholds of the partitioning process according to the value of spatial frequency measurement (SFM) is adopted to achieve the better image compression ratio.