Segmentation of Food Images by Local Extrema and GrabCut
Haruki Sugiyama, Gamhewage C. de Silva, Kiyoharu Aizawa · The Journal of The Institute of Image Information and Television Engineers · 2012
We developed a method to automatically segment meal images into food and non-food regions. In the proposed method, local luminance extrema are detected and GrabCut is applied to their convex hull. The recognition rates were 90.38% for food regions and 68.12% for non-food regions.