四元量化的综合鉴别函数滤波器实现比例不变相关识别

Quantization of Synthetic Discriminant Function Filter to a Quadrupole For Scale Invariant Pattern Recongition

设计一种基于综合鉴别函数(SDF)的比例不变相关识别滤波器,将频域中被SDF调制的相关输出函数进行复数形式的四元量化(QFCQ),并采用叠代算法调整SDF的系数,使输出的相关峰尖锐,区分能力强,同时满足不同比例图像等相关峰输出的准则,从而达到了比例不变识别的要求.计算机仿真实验对属于训练图像系列的目标图像进行了识别,并与基本的SDF及二元化纯相位SDF做了比较,结果表明该方法具有较好的识别效果.

A new scale-invariant correlation filter based on synthetic-discriminant-function （SDF） was designed. In this filter, the correlation output function modulated by SDF in frequency-domian was quantized to complexes of a quaduploe to obtain sharp correlation peak. To satisfy the equal correlation peak criteria , an iterative algorithm was utilized to modify the coefficients of SDF. Thus scale-invariant recognition was achieved. Computer simulation results show that the proposed method performs better compared with the basic SDF and the binary phase-only SDF（BPOF SDF） for the target images from training set.