Enhancement of Decompressed Images at Low Bit Rates

R. A. Gopinath?

IBM T. J. Watson Research, #J2-J05, 30 Saw Mill River Road, Hawthorne, NY 10542

M. Langy, H. Guozand J. E. Odegardx

Department of Electrical and Computer Engineering, Rice University, Houston, TX-77251

Paper presented and published in the Proceedings of SPIE, Mathematical Imaging: Wavelet Applications in Signal and Image Processing II (2303-30), July 27-29, 1994, San Diego, CA. This is also Technical Report Rice University, CML TR94-05

ABSTRACT

Transform coding at low bit rates introduces artifacts associated with the basis functions of the transform. For example, decompressed images based on the DCT (discrete cosine transform) - like JPEG16 - exhibit blocking artifacts at low bit rates. This paper proposes a post-processing scheme to enhance decompressed images that is potentially applicable in several situations. In particular, the method works remarkably well in deblocking" of DCT compressed images. The method is non-linear, computationally efficient, and spatially adaptive - and has the distinct feature that it removes artifacts while yet retaining sharp features in the images. An important implication of this result is that images coded using the JPEG standard can be efficiently post-processed to give significantly improved visual quality in the images.

Keywords: image enhancement, wavelet shrinkage, jpeg, image restoration, image compression

1 INTRODUCTION

Effective compression of still images such as the JPEG standard16 is important for many applications which use bandwidth-limited media such as radio channels or storage media with steadily growing yet limited capacity. Using these (lossy) compression methods it is possible to achieve compression ratios of 10 and more with only a small noticeable perceptual deterioration.16 However, for considerably higher compression ratios the perceptual quality decreases rapidly. There are essentially two undesired effects, the loss of image details and blocking. The former is a result of disregarding high frequency components during the quantization process. The latter is due to the DCT coding of 8 ? 8 blocks in the JPEG standard.

?This research was supported by BNR
yThis research was partially supported by the Alexander von Humboldt foundation and by AFOSR under grant F49620-1-0006 funded by DARPA
zThis research was partially supported by TI and by AFOSR under grant F49620-1-0006 funded by DARPA xThis research was supported by AFOSR under grant F49620-1-0006 funded by DARPA