Article directory
Image segmentation
- Divide the image into several disjoint regions.
- Classical digital image segmentation algorithms are generally based on one of two basic features of gray values: discontinuity and similarity.
Threshold-based, edge-based
-
Threshold-based: Computes one or more grayscale thresholds based on image grayscale features. Compare the gray value with a threshold, and finally classify the comparison results into appropriate categories.
-
Otsu method -
Edge-based: a collection of continuous pixels on the boundary line, which reflects the discontinuity of the local features of the image. Embodies sudden changes in image features such as grayscale, color, and texture.
Based on area, based on graph theory
- Region Segmentation:
- Region Growing Method
- Watershed Algorithm
- Graph theory segmentation
- Graph Cuts segmentation
- first take two seed points (foreground and background)
- then build a graph, the thickness of the edge on the way represents the size of the corresponding weight
- then find the combination of weight and minimum edge
- complete the function of image segmentation
- GrabCut segmentation
- object segmentation
Face Detection
Haar-like features + cascade classifier
Represents certain characteristics of Renlaiyun
pedestrian detection
HOG+SVM
DPM