Algorithms Of Image Line Feature Extraction Using Beamlet Transform

Beamlet transform is one of efficient tools to realize image multiscale geometric analysis. This paper analyze the basic theory of the continuous Beamlet transform and its applications in linear feature extraction in images. Beamlet is a frame in which line segments play an important role similar to the role played by points in wavelet analysis. Wavelet offer localized scale, location representation near fixed regions of space, Beamlet has localized scale location and orientation based on dyadically line segements. Line feature extraction based on Beamlet pyramid includes structureless algorithm, tree-based algorithm, local line segment chain algorithm and global line segement chain algorithm.In this dissertation, we investigate the algorithm to extract line and curve structures from grayscale images using Beamlet transform. This paper presents a method for line feature extraction based on Canny operator and Beamlet transform . First, Beamlet transform is applied to extract the line feature in images. Then, Canny operator is used to remove false line feature from the line feature extracted by Beamlet transform. The experiment results show that the new method is more effective. In addition, we also discuss a line feature detection algorithm which is based on arbitrarily shaped windows and Beamlet transform. The algorithm is realized in three steps. Firstly, we smooth noise in noisy image using arbitrarily shaped windows. Secondly, we fuse the images from different windows into a binary image. Finally, Beamlet transform is used to extract the line feature in the binary image. This method can extract the line feature well, including curvilinear feature. Experimental results indicate that, this method remains valid when the signal-noise-ratios are low.