Fast texture energy based image segmentation using Directional Walsh–Hadamard Transform and parametric active contour models

Texture image segmentation is an important issue in computer vision applications. Active contour models are one of the powerful tools that are able to detect and segment textured objects against textured backgrounds. However, problems concerning the speed of the contour convergence in the texture image have limited their utility. This paper presents a fast and efficient texture energy function in the parametric active contour models. In the proposed method, we apply a novel version of the Walsh–Hadamard transform, called the Directional Walsh–Hadamard Transform or DWHT for calculating texture features of the energy function. This DWHT-based energy function is fast and easy to implement, and hence suitable for real time applications. We will show that the proposed method can reduce the execution time, while maintaining close accuracy and consequently it is more efficient than the previous active contour based methods for texture image segmentation.