Analysis And Research On Image Edge Feature Extraction Algorithm

Image edge feature is one of the basic features of the image, including the most important information in the image. The image edge feature extraction is the basic of digital image processing, computer vision and pattern recognition. Currently, the digital image edge feature extraction has been widely used in the scope of target tracking, palm-print recognition, remote sensing image segmentation and has become one of the hotspots of digital image processing.Image edge feature extraction methods are divided into two types. One is based on the traditional image brightness gradient image edge feature extraction algorithm, which is characterized by the use of marginal changes in image brightness gradient extracted image. The other is based on the phase coherence frequency domain image edge feature extraction algorithm, which uses the fourier components of the image signal phase point as the edge of the most consistent feature point image.In this paper, the image edge feature extraction is research objectives, Based on the in-depth analysis of the traditional image edge feature extraction algorithms and phase coherence image edge feature extraction algorithm, we proposed the phase coherence multi-scale edge detection algorithm and pyramid multi-scale edge feature fusion algorithm. The main work and characteristics are as follows:(1) Traditional image edge feature extraction algorithm which is based on brightness gradient simulation experiments has been made, summarizes and analyzes the traditional algorithm has good effect in the absence of noise and edge, feature extraction based on first order differential operator, such as Roberts, Sobel algorithm affected by noise are small, and the edge feature extraction based on first order differential operator LoG affected by noise is large. Canny algorithm affected by noise is small.(2) System elaborated the basic principle and process optimization of phase consistency which is based on frequency domain, gives the detailed steps of phase consistency of edge feature extraction algorithm and simulation experiments, by compared with the traditional algorithms of image edge feature extraction, verify the phase consistency algorithm is not affected by the image brightness and contrast change.(3) Considering the image edge features has multi-scale, multi-resolution features, makes it hard to completely extract image edge feature from a single scale image, the multi-scale pyramid can carry on the multi-scale and multi-resolution representation of image. Therefore, this article presents a phase coherence multi-scale edge detection image edge feature extraction algorithm, which combine phase coherence method with Laplace multi-scale pyramid image representation method, which has certain advantages in compare with single scale phase consistency algorithm in dealing with noise and preserving edge details feature.(4) Considering the each phase coherence scale edge character cannot be combined for a simple sum, multi-scale edge feature fusion algorithm is proposed in this paper, which combine the large scale image provided location information with small scale images provided rich precise details, can get accurate edge feature information. By simulation experiments, through compared with phase consistency edge feature extraction algorithm, phase coherence multi-scale edge detection image edge feature extraction algorithm extract image edge feature is more integrity. Compared with other algorithms, the comprehensive performance improved.