1. What have the authors contributed in "Shape-based mutual segmentation" ?
The authors present a novel variational approach for simultaneous segmentation of two images of the same object taken from different viewpoints.. The evolving object contour in each image provides a dynamic prior for the segmentation of the other object view.. The suggested shape term incorporates the semantic knowledge gained in the segmentation process of the image pair, accounting for excess or deficient parts in the estimated object shape.
read more
![Figure 3: (a,b) Noisy (a) and corrupted (b) images of the same object taken from different view points. The initial contours are drawn in red. (c,d) Successful mutual segmentation results (red). (e) Superposition of the two images to demonstrate the misalignment. (f- h) Segmentation of each image by itself. The noisy image (g,h) was segmented twice with different weights of smoothness term: (g) The contribution of the smoothness term WLEN(t)(φLENt ) was restricted to [−1, 1] (refer to subsection 4.1 for details). The contour “mistakenly” follows image gradients that are due to noise. (h) The smoothness term WLEN(t)(φLENt ) was further stressed, i.e. its contributions were multiplied by two. The segmenting contour (red) is smoother but the gaps between the fingers are not well extracted.](/figures/figure-3-a-b-noisy-a-and-corrupted-b-images-of-the-same-3nxxck6o.webp)
