Multiresolution Morphological Object Detection

First,a pyramid representation of images is created using morphological wavelet transform. Second,a coarse-to-fine analysis scheme is implemented from the lowest-resolution to the ful-l resolution.At each resolution,watershed transform segments the image of the pyramid （or the previously obtained result） to generate a labeled image,which is relabeled by a region searching scheme.Experimental results with various kinds of image data demonstrate the algorithm’s efficiency,precision,and robustness to noise.