双正交小波弱目标提取

Haar小波边缘提取对弱小目标和噪声很难辨别,双正交小波继承了haar小波的优点,其滤波器系数比haar小波更长,所以双正交小波抗噪性强。提出采用双正交小波提取弱目标边缘,利用形态学滤波消除边缘提取后的孤立点噪声。实验结果表明:当目标对比度下降到一定程度时,sobel算子和haar小波无法提取出弱目标,此时双正交小波提取目标效果较好;对同一组图像序列采用不同算子边缘提取后,haar小波相关度在0.75~0.85,sobel为0.8~0.85,双正交小波为0.85~0.95,双正交小波目标提取稳定可靠,速度快。

Extraction of faint targets with biorthogonal wavelet

Haar wavelet is difficult to identify fain targets from noise under low signal-noise-ratio. Biorthogonal wavelet inherits the advantages of haar wavelet. Its filter coefficient is longer than haar wavelet so biorthogonal wavelet has more powerful anti-noise performance. The extraction of faint targets edge with biorthogonal wavelet and the elimination of isolated point noise by morphological filtering are presented in the paper. The experimental results show that when target contrast is reduced to a certain level, sobel operator and haar wavelet cannot identify faint target while the biorthogonal wavelet has a better target identification effect. Through comparison of a sequent of images from edge extraction by mean of haar, sobel and biorthogonal wavelet with original ones,the correlativity coefficient is as 0.75 ~0.85, 0.8~0.85 and 0.85~0.95 respectively.