Image/Video Compression


·       JPEG2000 image compression

·       Low bit rate coding

·       Tree-structured vector quantizer design


JPEG 2000 Image Compression

 

JPEG2000 is a new image compression standard. Being an active contributor to the JPEG2000 standard for a couple of years, I have been focusing on three different aspects of JPEG2000, i.e., visual optimization, palette-based JPEG2000 for graphics-like images, and special filter operation to reduce the tile boundary artifact of JPEG 2000 compressed images. Three of our proposals (one in each of these three aspects) have been adopted into the JPEG2000 standard.

 

The human visual system plays a key role in the final perceived quality of the compressed images. It is therefore desirable to allow system designers and users to take advantage of the current knowledge of visual perception and models in a compression system. We have developed a number of tools (adopted by JPEG-2000) that allow its users to exploit many properties of the human visual system (HVS) such as spatial frequency sensitivity, distortion-adaptive frequency sensitivity, color sensitivity, and the visual masking effects in the JPEG2000 framework. Our approach makes fully use of the HVS properties (including point-wise extended masking effect which is very adaptive) while still maintaining the scalability feature. As a result, the visual tool sets in JPEG-2000 are much richer than what is achievable in JPEG, where only spatially invariant frequency weighting can be exploited.  With such visual optimization tools, we can save bit rate by up to 50% for a comparable visual quality, when compared to MSE based R-D optimized approach. For more details, see Papers [1-4].

 

We also convinced the JPEG2000 standard committee to support palette-based coding to accommodate efficient compression of images with limited number of colors. We proposed a fast and efficient way for color re-indexing to maximize the compression performance of a palette-based JPEG2000 compression system. See Papers [5].

 

Realizing that typical tile-based JPEG2000 systems suffer from blocky artifacts at the tile boundary, we analyzed the cause of these blocky artifacts, and proposed a special filtering operation that can minimize the tile boundary artifacts. See Papers [6].

 

Papers:

 

1.        W. Zeng, S. Daly and S. Lei, “An overview of the visual optimization tools in JPEG 2000," invited paper, in special issue on JPEG-2000 of Signal Processing: Image Communication, Vol 17, no 1, pp. 85-104, Jan. 2002. (click here for preprint in pdf format)

2.        W. Zeng, S. Daly and S. Lei, "Visual optimization tools in JPEG 2000," invited paper, special session on JPEG-2000, Proc. IEEE Inter. Conf. Image Proc., Vancouver, Canada, 2000. (click here for preprint in pdf format )

3.        W. Zeng, S. Daly and S. Lei, "Point-wise extended visual masking for JPEG-2000 image compression," Proc. IEEE Inter. Conf. Image Proc., Vancouver, Canada, 2000.(click here for preprint in pdf format)

4.        S. Daly, W. Zeng, J. Li and S. Lei, "Visual masking in wavelet compression for JPEG2000," in IS&T/SPIE Symposium on Electronic Imageing: Image and Video Communications and Processing , vo l. 3974, San Jose, Jan. 2000. (click here for postcript file )

5.        W. Zeng, J. Li and S. Lei, "An efficient color re-indexing scheme for palette-based compression," Proc. IEEE Inter. Conf. Image Proc., Vancouver, Canada, 2000. (click here for preprint in pdf format )

6.        J. Wei, M. Pickering, M. Frater, J. Boman, J. Arnold and W. Zeng, “Boundary artifact reduction using odd tile length and the low pass first convention (OTLPF),” Proc. of SPIE Conf. on Applications of Digital Image Process., July 29 – Aug. 3, 2001. (click here for preprint in pdf format)

7.        W. Zeng, J. Li and S. Lei, "Adaptive wavelet transforms with spatially varying filters for scalable image coding," in Proc. IEEE Inter. Conf. Image Proc., Chicago, 1998. (Click here for postcript file)


Low Bit Rate Coding

The idea of incorporating an interpolator at the receiver has been used for block-based low bit rate coding to improve (about 15%-30% bit rate saving) over conventional JPEG/MPEG/H.26X that do not fully exploit the inter-block correlations. Similar ideas are also applicable to other applications, e.g., reducing computational (rendering) complexity in computer graphics. We also believe that incorporating information extracted by content-analysis of video into a compression/transmission system will result in improved visual quality.

 

Papers:

 

1.        W. Zeng and B. Liu, "Geometric-structure-based error concealment with novel applications in block-based low bit rate coding," IEEE Trans. on Cir. and Sys. for Video Technology, pp. 648-665, June 1999.  ( 49 citations using Google Scholar search)

2.        Wenjun Zeng and Bede Liu, "Directional spatial interpolation for DCT-based low bit rate coding", Proc. IEEE International Conf. Acoustic, Speech, and Signal Processing, Atlanta, May 1996.

 


 

Last Updated :