Localized Video Compression for Machine Vision
A three-dimensional vector quantization system is introduced suitable for video compression. The basic characteristics of slow or repeated scenes in robot vision are used as the basic assumptions of the proposed approach. Accordingly, the localized history of the sequence is used to create localized codebooks, thus representing current visual information as transformed versions of previous details. The results indicate a high compression ratio with high quality of the perceived sequence. The structure of the algorithm is mostly parallel, making it suitable for efficient hardware implementation.
Unable to display preview. Download preview PDF.
- 1.IEEE Trans. Image Processing, Special Issue on Sequence Coding, (September 1994). 278Google Scholar
- 2.C. S. Choi, H. Harashima, T. Takebe: Analysis and synthesis of facial expressions in knowledge-based coding of facial image sequences. IEEE ICASSP (1991). 278Google Scholar
- 4.C. I Podilchuk, N. S. Jayant, P. Noll: Sparse codebooks for the quantization of nondominant sub-bands in image coding. IEEE ICASSP, (1990) 2101–2104. 278Google Scholar
- 5.G. R. Giunta, T.R. Reed, M. Kunt: Image sequence coding using oriented edges. Image Communication, 2 (1990) 429–440. 278Google Scholar
- 7.J.W. Woods, S.D. O’Neil: Subband coding of images. IEEE Trans. on Signal Processing, ASSP-34 (1986) 1278–1288. 278Google Scholar
- 13.M. Goldberg, H.-F. Sun: Image sequence coding by three-dimensional block vector quantization. IEEE Proceedings, 133, Pt. F, No. 5 (1986) 482–486. 282Google Scholar