MPEG-4 Still Image Coding for Noisy Images

Dr. Daniel Lun and Mr. C.L. Chan

In this project, a fine-tuning technique is developed for the MPEG-4 still image codec so that when encoding noisy images, better visual quality can be achieved with much lower bit rate.  More specifically, the fine tuning technique increases the compression rate of the MPEG-4 still image codec by about 30% with decoded images having similar quality as the original ones.  The advantages of this approach include the following: (1) It requires only a slight modification to the encoder such that the computational complexity, hence the computation time, the same as before. (2) It does not require any modification to the decoder.

Original decoded noisy image Bit rate = 0.613bpp PSNR=29.42	Fine-tuned Result Bit rate = 0.613bpp PSNR=29.42	Original decoded noisy image Bit rate = 0.613bpp PSNR=29.42	Fine-tuned Bit rate = 0.613bpp PSNR=29.42