Edge detection using wavelets

Abstract

This paper studies the edge-detecting characteristics of the 2-D discrete wavelet transform. Our problem is to automatically detect edges. Since a common claim about the wavelet transform is that it splits images into an approximation and details, which contain edges, we use it in our experiments. First, to determine its efficacy, the 2-D discrete wavelet transform is compared to other common edgedetection methods. Also, a number of combinatorial methods for the octaves are examined in the comparison. Due to this work, a novel boundary-tracing algorithm is developed, to follow edges around an object of interest.

A novel framework for compressed sensing brain MRI edge detection and de-noising

“…These methods lack the ability to surpass impulsive noise, and are incapable of recognizing slow varying intensities. Wavelet methods [45], [58], [9], [62], [2], on the other hand, utilize the natural behaviour of wavelets to divide a signal into different frequency bands: low frequency (approximation), and high frequency (in 3 different orientations). In addition, wavelets are highly localized and well capable of recognizing discontinuities in a signal, and respond better to impulsive noise components in a signal.…”

Comparison of Traditional Approach for Edge Detection with Soft Computing Approach

“…Wavelet transform [21] and mathematical morphology with fuzzy rules [6] is used in proposed method. Multi-scale wavelet transform is used for smoothing the image.…”

Discrete Wavelets on Edges

“…Recent works still use the idea of convolution by a family of Gaussians (Sumengen & Manjunath (2005), Zhang et al (2010)) and nonlinear diffusion filters (Tremblais & Augereau (2004)). Other works are wavelet-based, as can be seen in (Belkasim et al, 2007), (Shih & Tseng, 2005), (Han & Shi, 2007), Brannock & Weeks (2006) and Heric & Zazula (2007). Sumengen & Manjunath (2005) create an Edgeflow vector field where the vector flow is oriented towards the borders at either side of the boundary.…”

“…Once a decomposition level is chosen, textures are then removed from the original image by the reconstruction of low frequencies only. The problem for Brannock & Weeks (2006) is to automatically detect edges. To determine its efficacy, the 2D discrete wavelet transform is compared to other common edge detection methods.…”