为了改善JPEG算法的编码性能,提出了一种新的基于自学习特征映射算法(SLM)的静止图像编码方案,基本思想是:1)先用矢量量化(VQ)对图像进行预测,然后进行DCT编码;2)采用新的SL,M算法来改善vQ码书性能.实验表明,与JPEG算法相比,该方案具有更高的压缩比和更好的图像质量.
To improve the performance of JPEG algorithm, a new scheme for still image coding based upon self-learning neural network is presented. It uses vector quantization(VQ) based upon self-learning feature maps(SLM) to predict input image, then does DCT coding. Experimental results show that this scheme is better than JPEG algorithm.