This dissertation presents a new approach to wavelet-based edge detection of noisy images by using multiresolution analysis. We used the discrete wavelet transform with the correlation of noisy wavelet coefficients between different scales to remove noise. In this research, experiments in 1-D and 2-D were performed to test the noise performance of the new technique. Experimental results were analyzed and compared with the 1st and 2 nd derivative of edges. The new method showed better performance than other methods in the presence of noise. |