Entropy-based spatially-varying adjustment of active contour parameters

Abstract

Parameter adjustment is a crucial, open issue in active contour methodology. Most state-of-the-art active contours are empirically adjusted on a trial and error basis. Such an empirical approach lacks scientific foundation, leads to suboptimal segmentation results and requires technical skills from the end-user. This work introduces a method for automatic adjustment of active contour parameters, which is based on image entropy. In addition, instead of being uniform, the parameter values calculated are spatiallyvarying, so as to reflect textural variations over the image. Experimental evaluation of the proposed method is conducted on thyroid US images, liver MRI images, as well as on real-world photographs. The results indicate that the proposed method is capable of identifying plausible object boundaries, obtaining a segmentation quality which is comparable to the one obtained with empirical parameter adjustment. Moreover, the applicability of the proposed method is not confined on a single active contour variation.

Superpixel-based and boundary-sensitive convolutional neural network for automated liver segmentation

“…Given a superpixel map, we also measured the information contained within each superpixel by computing a saliency map based on Shannon entropy (Mylona et al , 2012), as figure 3(c). Here, we selected the superpixel with the mean entropy value greater than 1.5 to be the training patches.…”

Video retargeting through spatio‐temporal seam carving using Kalman filter

“…The first step in the system is to compute energy maps of video frames, which quantify the importance of each pixel in a frame. Measures such as gradient magnitude [10,11], saliency [43], salient region shapes [34] or local entropy [44] can be used the energy or importance map computation. As our main focus is on computing spatio-temporal seams given any kind of energy maps, irrespective of how they are computed, we consider the use the gradient magnitude as it is the most popular one.…”

Selection of parameters in active contours for the phenotypic analysis of plants

“…Allili et al (2008) propose a Bayesian model where the parameters are treated as hyper-parameters to be estimated. Mylona et al (2012) and Zhu et al (2011) also propose methods to dynamically update parameter values. However, their techniques update parameters spatially, selecting different values for different regions of the evolving curve.…”