I want to put up the picture first. This is drawn after research and study, although the effect is not very good.
Below is how I converted the original image into HSV and RGB channels, and then performed Graph Cuts on each of their single channels.
1.Graph Cuts is a method in image segmentation, which solves the energy function by using the form of a graph. To determine the energy function, the corresponding weights can be assigned to the edges of the graph, so that the energy function can be completely converted into an S/T graph, and the maximum flow/minimum cut of the graph can be solved by using the properties of the graph to find the global optimum. global energy minimum.
2.graph cut is a variety of segmentation algorithms based on graph (graph theory), such as NCut, minimum cut, RatioCut, graph cut, etc. graph cuts is a graph cut, a method used to solve the energy equation based on MRF.
Disadvantages:
The result of the second-class segmentation is that the global energy function is the smallest, the global optimal solution, the multi-class problem cannot be optimal, and can only be a local optimal solution; there are
advantages in processing images with obvious differences in pixel values, high-contrast images,
The processing image contains noise or occlusion and the segmentation effect is poor, and it is easy to get wrong segmentation;
the user needs to manually mark some foreground and foreground pixels, and there is manual intervention;
the difference between the initial mark and the foreground pixels will lead to different degrees of final segmentation effect, graphcuts algorithm Very sensitive to user interaction;
when the image contains a complex background, more user interaction is required, using a pen/brush to mark more background and foreground pixels.
Code download: GitHub - DamonZCR/GraphCut: Image Segmentation Using Maximum Flow Minimum Cut
