Edge Detection Based on Regulated Morphological Gradient

摘要

Morphological gradient transform is a common way to detect both sharp and transition edges, but it tend to result in thick edges. A regulated morphological gradient method for edge detection is proposed corroding to Soille multi-scale gradient, while opening by reconstruction top-hat is used to replace traditional top-hat and ultimate erosion to erosion. First, morphological thick gradient is calculated, and then top-hat by reconstruction and ultimate erosion are employed to extract thin edges with less essential edges lost, and to avoid the small space edges merging. Experiments show that this method can efficiently detect thin edges without essential edges lost and edge point position bias under noiseless case, also it can result in good edges in the noise corrupted condition when the pre-filter is employed.