A Fusion Edge Detection Method Based On Improved LOG Operator And Wavelet Transform

In the methods of boundary segmentation, edge detection is a basalprocessing method. Wavelet analysis as a powerful tool studying non-stationarysignal has proven to be an important means for edge detection. In addition, thereare some commonly used classical edge detection operators. In this paper, thetraditional LOG operator and wavelet transform are studied respectively. Specificwork is as follows.Firstly, the traditional LOG operator does not have the adaptive ability ofselecting the variance of Gaussian filtering. The denoising and the edgepreserving effect are affected directly by the variance. And when the imagecontains a lot of noise, as the LOG operator is the second-order derivativeoperator and very sensitive to noise, therefore, it is difficult to distinguisheffectively the edge points and the noise points. For this reason, the LOGoperator has been improved by the grayscale morphology filtering replacing thegaussian filter, which can reduce effectively the noise, but also maintain the edgestrength and details.Secondly, the result of the traditional wavelet transform detecting the imagewith pepper and salt noise is not quite good, which almost can not remove thesalt and pepper noise. With regard to this, wavelet transform has been combinedwith morphological structure elements of different scales. Not only is pepper andsalt noise removed by using improved wavelet transform, but also the edge iscontinuous and more information is contained.Finally, based on improved LOG operator and wavelet transform combiningwith morphology, an edge detection method of image fusion is presented in thispaper. The edge of the original image is extracted respectively by the two methods, and then the gained edges make fusion. It is shown that the advantagesof the two kinds of improved algorithms are focused on the fusion image, in thefinal, the effect of the complete edge and the accurate positioning is achieved.