摘要
为了提高图象信号处理水平,可以采用一种新的多分辨率、多方向的图象变换方法。该变换采用非分离的钻石形和扇形滤波器进行方向滤波,利用统一的采样矩阵进行样本减(增)采样和方向重排,通过级联双通道树型结构得到多方向的频谱分割。该方法使每层分解后的图象具有一个低频子带和十二个高频子带,这些子带都为规则的方形矩阵,可以通过对低频子带的迭代分解实现多分辨率分析,具有临界采样和理想重构性质。
In this paper, a new multiresolution and directional transform had been proposed. The new transform is achieved by cascading the two channel tree structure, which adopted nonseparable diamond and fan filter to realize frequency partition and utilized uniform sampling matrices to rearrange and change data rate. The im- age can be deeom~ into one lowpass subband and twelve highpass subbands in each level. Besides, each ooef- fieient subbands is regular square matriee. The multiresolution representation can be achieved by simply reiterating the transform on the low-pass subband, and it has the maximally decimation and perfect reconstruction property.
出处
《陕西教育学院学报》
2010年第1期102-106,共5页
Journal of Shaanxi Institute of Education
关键词
多分辨率
方向滤波器组
非分离滤波器
采样矩阵
麦克劳琳变换
multiresolution
directional filter banks
nonseparable filter
resampling matrix
McClellan transiorm