Computer Vision News - June 2016

Top Ranking Methods The two successful participant teams suggest different approaches. The top ranking team suggests a method which first detects the nuclei within each cervical cell volume by using an iterative thresholding approach which looks for the darkest pixel in the image (which usually corresponds to the nucleus). It then segments cell clumps (composed by overlapping or isolated cells) via a Gaussian mixture model with two components (foreground and background), learned based on pixel intensities. Finally, it segments the cytoplasm of individual cell’s within each cell clump. The team’s paper details the operation used when more than one nucleus is found inside a cell clump. Computer Vision News Challenge 29 Dear reader, How do you like Computer Vision News? Did you enjoy reading it? Give us feedback here: It will take you only 2 minutes to fill and it will help us give the computer vision community the great magazine it deserves! Give us feedback, please (click here) FEEDBACK The second team suggests a model based on three steps: first, a rough segmentation of subcellular compartments using Super Pixel combined to Voronoi Diagrams (SPVD) ; the algorithms are intertwined to deliver an efficient pipeline for automated nucleus and cytoplasm segmentation. Second, algorithms using calculus of variations construct edge maps. Third, mathematical methods such as morphological reconstruction determine minimum enclosing ellipse to deliver a refined cytoplasm boundary. The details of this model are here . Both new methods successfully allow cervical cell detection and individual segmentation of nucleus and cytoplasm. Squamous metaplastic cells from pap smear