基于平均偏差的快速分形图像编码

Fast fractal image encoding based on mean deviation

针对基本分形编码方法耗时过长的不足,提出了一种基于平均偏差的快速算法。首先对所有父块按平均偏差排序,然后将子块按标准差分为平滑块和非平滑块,对于平滑子块直接存储其均值;对于非平滑子块,在以其平均偏差意义下的最近邻父块为中心的邻域内搜索最优匹配父块,并且引入误差阈值控制子块的搜索过程。实验表明,较之基本分形编码算法,所提快速算法的编码速度大大提高,并且优于基于形态特征的快速算法。

One shortcoming of the baseline fractal algorithm is the excessive cost of time. A fast fractal encoding algorithm based on mean deviation is proposed. First, all the domain blocks are ordered by their mean deviation. Then, every range block is classified into smooth or non-smooth block according to its root mean square error. The mean is kept for every smooth range block. Every non- smooth range block searched for the best domain block in the vicinity of the nearest domain block in the sense of mean deviation, andan error threshold is used to control the searching area. The experimental results show that the proposed algorithm is much faster than the baseline fractal algorithm and better than the algorithm based on shape feature.