|
JPEG has become the industrial standard for still image
compression. At medium bitrate, JPEG performs well. At low
bitrate, on the other hand, JPEG suffers from serious blocking
effects. In this thesis, we present a JPEG-based coding system.
By adding pre-/post- processing on JPEG, a simple and easy to
implement system is constructed. The pre-/post- processing
schemes employ two techniques - classification and subsampling.
An input image is segmented into high/low activity blocks by
combining the L1 norm and the mean difference method. Low
activity blocks are subsampled and encoded by JPEG, while hige
activity blocks are directly encoded by JPEG. As a result, it
yields better, in comparison with JPEG, quality in both smooth
and complex areas of image at low bitrates. Simpler schemes by
applying only classification technique or subsampling technique
are discussed also.
|