利用余弦变换实现图像压缩：

DCT先将整体图像分成N×N像素块（一般N=8 ，即64个像素块），

再对N×N块像素逐一进行DCT变换。由于大多数图像高频分量较小，

相应于图像高频成分的失真不太敏感，可以用更粗的量化，在保证所要求的图质下，

舍弃某些次要信息。-Cosine transform image compression: DCT NN first image is divided into blocks of pixels as a whole (usually N = 8, i.e., 64 pixel blocks), and then each of the NN pixel block DCT transform. Since most high-frequency component image is small, the high frequency component corresponding to the image distortion less sensitive, more coarse quantization can be used, quality assurance in FIG required to discard some of the secondary information.