摘要
由于基本分形编码算法耗时较长,因此提出采用边缘分类和平均偏差比较的分形编码算法。按边缘分类算法将父块分为边缘类和非边缘类,各类父块按平均偏差排序。对每一子块,计算其标准差,若小于设定的阈值,则取其均值作为分形码,否则按边缘分类算法确定其类别,并在同类父块库中找到其平均偏差意义下的最近邻父块,然后在该父块的指定邻域内搜索满足误差阈值要求的父块。实验结果显示该算法优于文献[1-3]的算法。
Since basic fractal encoding algorithm consumes longer time, we present a new algorithm which uses edge classification and mean deviation comparison. The parent blocks are classified to the edge class and non-edge class according to the proposed edge classification algorithm, and the parent blocks of each class are sorted according to their mean deviations. Every block will be calculated its root mean square error （ RMSE）, if less than the given threshold, its mean value will be taken as the fractal code, otherwise its class will be determined according to edge classification algorithm, and its nearest neighbour parent block in the sense of mean deviation will be found out in the base of same-class parent blocks, then the parent block satisfying the error threshold is to be searched within the specified neighbourhood of the parent block. Experimental results demonstrate that the proposed algorithm outperforms the algorithms proposed in literature [ 1 ], [ 2 ] and [3].
出处
《计算机应用与软件》
CSCD
2015年第2期211-214,共4页
Computer Applications and Software
关键词
分形
图像编码
边缘
平均偏差
标准差
Fractal Image encoding
Edge
Mean deviation
Root mean square error
