Most video compression techniques do not operate directly in the highly-correlated RGB space. Instead, they transform the primary colors into a decorrelated color space, such as YCbCr or YIQ, where the compression takes place. In this paper we propose a different approach. The video stream is sub-divided into 3D blocks (x, y, t), of high correlation between the color components. In each block, we encode two of the color components as polynomials of the third component (defined as `base color'). The base color is then compressed using conventional tools, while the two remaining colors are transmitted to the decoder by their polynomial coefficients. Experimental results show that the proposed approach outperforms presently available methods. Our conclusion is that the high correlation between primary RGB colors could be helpful for video coding and that the new spatio-temporal approach to video compression is more efficient than conventional decorrelation-based techniques.
展开▼
