Application Of Wavelet Neural Network Feature Extraction Technique In Image Edge Detection

The edge contains most of the information of an image, which givesthe contours of the image. And edge detection is the foundation forother image processing technology, such as image segmentation, imagecompression, and pattern recognition and so on. This thesis, which isabout edge detection technology for grey image, has done thefollowing work:1、Introduced the image edge and classification, summarized theedge detection methods, discussed the classic edge detection operatorin detail.2、Provided an initial introduction to the wavelet transform and neuralnetwork theory, studied how to combine the characteristics of them toconstruct wavelet neural networks(WNN), researched and analyzed thestructure and classification of WNN and its BP learning algorithmsystematically.3、Carried out comparison and analysis on the characteristics ofvarious classic edge detection operator through experiments, pointedout the advantages and disadvantages of them and their respectiveapplication scope: the algorithm of the edge detection operator basedon gradient is simple, but its anti-noise performance is poor, so it can beused for these image whose edge are clear; the linear filtering edgedetection operator can be applied to those image whose backgroundare more complex.。The experiment results showed that its detectionperformance is better than gradient operator edge detection method.4、Realized WNN for edge detection via programming. The resultsshowed that, WNN have a good performance of detecting continuousweak edge in noise images, while the classic operator cannot do that.But when detecting standard mage, WNN do not perform well.The WNNof this thesis has a property of strong robustness. The work of this thesis has provided a certain reference value for theresearch and application of edge detection method in imageprocessing.