Search

A complex domain-based approach of active contour model has been proposed for image segmentation which deforms iteratively to partition an image into various useful regions. A new region-based complex pressure force function has been designed which gives shape to the dynamic active contour using image forces, and efficiently control the propagation of moving interface. This model makes the level set function binary, uses Gaussian smoothing kernel to regulate and avoid re-initialisation procedure. The working scheme of the model is as: the real image data is converted into complex data by iota (i) times and the average iota (i) times of horizontal and vertical components of gradient is inserted in the model to develop complete complex gradient. The proposed model can be implemented by simple finite difference scheme. The efficiency and robustness of the authors model have been verified.

A novel region based Active Contour Model (ACM) for image segmentation is presented using image local information for intensity inhomogeneity images. A transcendental (trigonometric) energy functional based on Local Fitted Image (LFI) energy is suggested to extract the image local information. The difference between the original and fitting image is introduced as an angular constraint of the trigonometric function. This new approach has the ability to handle images with severe intensity inhomogeneity, multiple intensity classes and regions with non-homogeneous pixel intensities. To regularize and maintain the level set function, Gaussian filtering procedure is adopted that not only maintains the level set consistency but also eliminates re-initialization procedure to avoid computational complexity. Furthermore, the proposed model holds boundary regularization property and achieves sub-pixel accuracy. Experimental results on several real and synthetic images demonstrate that the proposed model has higher performance and better accuracy in comparison to the state-of-the-art models.