中文摘要：

针对卡洛变换（K-LT）应用于彩色图像压缩时因额外码流开销大而导致编/解码速度慢的缺点,提出了一种最大限度地减少额外码流开销的彩色图像压缩方法：先将RGB色彩空间的图像转换到YCbCr色彩空间采样,初步去除各颜色分量间的相关性;然后合并各色彩分量,进一步利用分块K-LT去除分量内部的信息相关性;最后对K-LT的重要系数采取查表的策略进行量化和哈夫曼编码.分块K-LT大幅度降低了向量空间的维数及变换矩阵的维数,减少了计算量和额外码流的存储开销;而查表提取K-LT重要特征系数的策略进一步节省了码流的存储空间.实验结果表明,该方法不仅编/解码速度快,而且相同压缩比下的峰值信噪比明显高于JPEG2000压缩方法.

英文摘要：

Aiming at the problem that when the Karhunen-Loeve transform （K-LT） is applied to color image compression,the overhead of additional code stream brings about a slow encoding/decoding speed,a color image compression method for reducing the overhead of additional code stream to a minimun is presented.The method takes the steps below：First,an image in the RGB color space is transformed into the YCbCr color space for sampling and preliminary removal of the corelation between each color component.Then,these color components are again merged so as to further make use of the block K-LT to remove the internal information relevance.Finally,the important characteristic coefficients of the K-LT are quantified and encoded in the Huffman way by adopting the look-up-table strategy.The block K-LT significantly reduces the dimension of the vector space and the transformation matrix,so it reduces the amount of calculation and the store-overhead of additional code stream; on the other hand,the look-up-table strategy saves further on the store-space of code stream.The experimental results show,in the case of the same compression ratio,that the method＇ s encoding / decoding speed is not only faster,but also the peak signal to noise ratio is significantly higher than that of the JPEG2000 compression method.