An image segmentation algorithm based on WT-MMGR superpixel technique
In the field of image processing,spectral clustering algorithms are widely used in image segmentation tasks.How-ever,for images with complex backgrounds and unclear color boundaries,traditional spectral clustering algorithms often cannot ac-curately segment target objects and require a large amount of computing resources and time.Therefore,an improved spectral clus-tering algorithm based on multi-scale morphological gradient reconstructed watershed superpixels is designed.It uses the multi-scale morphological gradient reconstructed watershed superpixel algorithm as a preprocessing step of the spectral clustering algorithm.The processed results are used as the input of spectral clustering to achieve the computational purpose of the spectral clustering algorithm.The experimental results of the verification algorithm show that the proposed improved spectral clustering al-gorithm has better edge fit in the image segmentation task,and the running speed of the program has been significantly improved.
万方数据
