Application Of Beamlet Transform In Line Feature Extraction From Images

In order to efficiently represent two-dimensional data such as grayscale images, multiscale geometric analysis have been developed recently. Beamlet transform is one of efficient tools to realize multiscale geometric analysis. Beamlet transform represents signals using line segments as basis elements, unlike wavelet transforms that use'point'as basis elements.In the dissertation, we investigate the algorithm to extract line and curve structures from grayscale images using beamlet transform. A three-stage algorithm based on beamlet transform is proposed to extract image. First, the directional filtering using a set of directional windows is used to highlight the line and curve features along different directions. Secondly, the thresholding process is used to transfer the obtained grayscale images to a set of binary images and these binary images are fused into a single binary image by the logical'or'operator. Finally, beamlet transform is applied to the fused binary image to extract the line and curve features in the original noisy image. In addition, we also give a line feature extraction algorithm using jointly wavelet transform and beamlet transform, where wavelet transform serves for enhancing image's features while beamlet transform is used to extract these line features in images. The experimental results show that the proposed algorithms can achieve satisfactory performance.