Curvature analysis using multiresolution techniques

Type Thesis

Rate 0.0 /5

Details

Files

Reviews

Download

Summary

Two computationally efficient and simple to implement schemes for detecting corners are proposed. The first scheme is based on Gabor filtering and the second is based on wavelet transform. The first technique detects only the corners and does not require preprocessing. This technique is suitable for both binary and gray level images and is suitable for both hot and cold objects. It is particularly useful when the image contains multiple objects and/or objects under occlusion. The second scheme detects corners, inflection points as well as smooth joins. This technique was also tested with various well known wavelets and the results are summarized. The two schemes have been successfully tested under noisy conditions. They were compared with currently existing techniques to demonstrate their competitiveness and computational efficiency.