Adaptive Edge Detection Based On Wavelet Square Multiple Scales Products

Edge detection is one of the vital technologies in image segmentation, image compression, visual tracking etc., concerned by many researchers. Therefore, many algorithms have emerged. However, in practice, due to the influence of noise, the detection performance is restricted. In this paper, we focus our research on improving the localization performance and enhancing noise resistance of the image edge detection algorithms.This paper proposes an edge detection method based on wavelet square multiple scales products, utilizing the spatial correlation between wavelet transform coefficients of signal and noise at multiple scales. We define a scale product function and multiply the wavelet coefficients together between adjacent scales to magnify the edges and suppress the noise. Experimental results show that this method has better localization performance.In the algorithm, considering the low SNR environment, we introduce an adaptive scale factor in scales products domain to measure the relative intensity of the signal and noise, and overcome the impact of noise on the detection performance.By comparing the experimental results, in low SNR environment, our method is able to achieve more than 93% positioning accuracy on standard test image. In high SNR environment, our method and other method based on wavelet multiple scales perform similarly. However, in low SNR environment, our method can improve SNR about 0.1d B, and PSNR can improve about 1d B compared with other method based on wavelet multiple scales. It shows that our algorithm is robust to noise.