Visual information transmission model for image coding based on the orientation selectivity of neurons in the visual cortex
✍ Scribed by Yasunari Yokota; Yasuhiko Masuda; Masae Yokota
- Publisher
- John Wiley and Sons
- Year
- 2000
- Tongue
- English
- Weight
- 920 KB
- Volume
- 31
- Category
- Article
- ISSN
- 0882-1666
No coin nor oath required. For personal study only.
✦ Synopsis
There have been proposed several image coding methods based on the orientation-based information processing mechanism of the visual system. There has been little analysis, however, of the effects of the statistical properties of the object of coding, or the orientation-based information processing mechanism, on the coding performance. This paper considers the application to image coding, and shows that the orientation-selective processing mechanism is effective in reducing the transmission rate in the visual system. A visual information transmission model is proposed in which the visual information is classified according to orientation, and each orientation class is transform coded by the orthogonal transformation prepared for each orientation. In such a transmission model, the transmission rate cannot be reduced, so long as the input to the visual system follows the Gaussian distribution, compared to the usual transform coding using a unified orthogonal transformation. In contrast, this paper considers an image which is the actual input to the visual system, and it is shown by an example that the transmission rate can be reduced by the orientation-selective processing mechanism, compared to the usual transform coding. It is further shown that the improvement is more remarkable when the visual input has more diversified local orientations. These results suggest that the orientation-selective processing mechanism can be applied to image coding.