一种基于四维离散余弦变换的视频编码方法

为有效地去除Y,U,V3帧之间内在的相关性,提出了一种基于四维离散余弦变换理论的彩色视频压缩编码方法。该方法将彩色视频放在四维数学模型中考虑,将彩色视频的图象宽度、图像高度、彩色空间的3帧（Y,U,V）,以及视频序列的连续3帧分别看作四维数学模型中的“四维”,利用四维离散余弦变换去除彩色视频中图象的相邻像素之间、Y,U,V3帧之间、以及视频序列相邻帧之间的相关性,并结合矢量量化进行压缩。实验结果表明,利用该方法对“Miss America”彩色视频进行压缩编码,在同等信噪比的条件下,可获得比标准MPEG-2压缩编码方法约4倍的压缩比。

Video coding technique based on four-dimensional discrete cosine transform

In traditional color video compression coding method, the color space of R, G, B is translated into Y, U, V firstly, then the three frames of Y, U, V are coded using the same method respectively. Thus, the inherent correlation among Y, U and V frame can not be wiped off effectively. A color video coding technique based on four-dimensional discrete cosine transform theory is presented. The color video is considered in the four-dimensional mathematical model, and the image width, image height, three frames (Y, U, V) of the color space, and three continuous frames of video sequence are regarded as the four dimension of the four-dimensional mathematical model. Further, correlations among adjacent pixels in an image, Y, U and V frame in the color space, adjacent frames in a video are all wiped off by four-dimensional discrete cosine transform. Vector quantization is used to compress the transformed coefficients. Experimental results show that color video compression for Miss America using the proposed method can approach four times of compression ratio than that of using standard MPEG method under the similar PSNR.