Computer Vision News - March 2017

Results: The authors used two functions to compare and evaluate their method: one, developed by Zhou et al in 2010, based on the Hamilton-Jacobi approach - marked IVC2010; the other, developed by Wang et al in 2013, based on the Mumford-Shah approach - marked IVC2013. On the left is the input image, on the right the ground truth segmentation. The following figure presents the new level set formulation segmentation (on the right), side by side with the segmentation results of IVC2010 (left) and IVC2013 (middle). It can be seen that the new formulation result is the closest to the ground truth indicated. The following figure shows the results of liver tumor segmentation by using the new level set formulation - the red contours. The yellow contours denote fuzzy clustering initialization, while the green contours denote ground truth manual delineation by medical professionals. The results can be seen to be relatively close to the ground truths indicated. Computer Vision News Research 29 Research