Active contour for noisy image segmentation based on contourlet transform

Abstract

Active contour is one of the most successful variational models in image segmentation, pattern analysis, and computer vision. However, traditional active contour models not only require much expensive computation but are very sensitive to noise. We propose a scheme for noisy image segmentation integrating the active contour model with the contourlet transform, an optimal sparse representation of an image. Having reconstructed all the scale maps, we downsample the last but one scale map twice. Then, we apply the active contour model on the coarsest scale map and take the segmentation results as the initial curves for the finer scale map. Experiments have demonstrated that our proposed method can yield desired segmentation results both in real and synthetic images.

A Novel Approach for Segmentation of Brain Image using a Multiscale Transform and a Region Based Active Contour

“…The Curvelet transform is a multi scale transform with better directional sensitivity .It helps to extract the image detail at various scales and directions according to the features of interest to be extracted. Integrating a multiscale , multiresolution transform such as Curvelet transform with Active contour models can effectively and intuitively solve those problems of Noise inhomogenity [13].The proposed Multiscale active contour segmentation model uses the entire scale space, to introduce the geometry of multiscale images in the segmentation process. The extracted multiscale structures will efficiently improve the robustness and the performance of standard shape analysis segmentation techniques such as shape recognition and shape registration and is able to extract convex and concave object based on coarse-to-fine scale and small-to-big size strategies.…”

“…This is known as segmentation [1]. The concept of Active contour (AC) models for segmentation the original model were initially proposed by Kass et al [13].This classical approach drives an initial contour towards the boundaries of the objects by minimizing an energy function whose minimum is obtained at the boundaries of the object(s). Chanvese active contours [26] establish the most robust and efficient method of image segmentation than the classical methods of histogram, thresholding, gradient based methods etc.…”

“…Combining a multiscale, multiresolution transform such as Curvelet transform with Active contour models can effectively and intuitively solve those problems of Noise inhomogenity [13]. Firstly, with the coarse-to-fine scale and small-to-big size strategies, the Activecontour models cost less expensive for computation because the rough segmentation results of the coarse scale maps can be taken as the initial contour of the following scale map.…”