Design of Linear-Phase Mirror-Image Symmetric Filter Banks for Image Compression Coding

The discrete cosine transform (DCT) has widely been used in image and video compression coding. However, the conventional DCT causes severe block artifacts in the compressed image in low-bit-rate coding applications. The block effect is perceived as visible discontinuities cross the block boundaries. Many researchers have proposed other techniques, such as the filter bank approach, to remove or suppress this undesirable feature. In this project with Algolith, a leading image enhancement solutions company, a new method for the design of linear-phase mirror-image symmetric filter banks will be developed based on the modification of conventional DCT. A lattice structure that is robust to quantization error and has a minimal processing delay will be derived for fast implementation of the filter bank. The lattice structure will then be optimized using the semi-definite programming method. The developed linear-phase mirror-image symmetric filter bank in the form of lattice structure will be used for image compression coding to reduce the block artifact and boost the coding gain.

Chao Wu
Faculty Supervisor: 
Dr. Wei-Ping Zhu