대한전자공학회:학술대회논문집 (Proceedings of the IEEK Conference)
- 대한전자공학회 2000년도 ITC-CSCC -1
- /
- Pages.54-57
- /
- 2000
New Distortion Measure for Vector Quantization of Image
- Lee, Kyeong-Hwan (School of Electronic and Electrical Engineering, Kyungpook National Univ) ;
- Park, Jung-Hyun (School of Electronic and Electrical Engineering, Kyungpook National Univ) ;
- Jung, Tae-Yeon (School of Electronic and Electrical Engineering, Kyungpook National Univ) ;
- Kim, Duk-Gyoo (School of Electronic and Electrical Engineering, Kyungpook National Univ)
- 발행 : 2000.07.01
초록
In vector quantization (VQ), mean squared difference (MSD) is a widely used distance measure between vectors. But the distance between the means of each vector elements appears as a dominant quantity in MSD. In the case of image vectors, the coincidence of edge patterns is also important when the human visual system (HVS) is considered. Therefore, we propose a new distance measure that uses the variance of differences to encode vectors and to design codebooks. It can choose more proper codewords to reduce edge degradations and make a useful codebook, which has lots of various edge codewords in place of redundant shades.
키워드