A Extraction Method for Image Significant Region Based on Improved Region Growing
Danhua Liu · Modern Computer · 2012
Proposes an improved region growing method to extract significant region.Firstly,different from the pixel-based or the simple N×N region based region growing method,uses the watershed transform based on morphology for the initial segmentation of the original image.Designs the algorithm of automatic seeded region selection based on local saliency and relative position by using visual attention mechanism.In the region growth process,presents a hybrid region dissimilarity function to measure the degree of similarity between two regions.It combines the relative boundary integrity criteria,the boundary length criteria with the traditional region homogeneity criteria,increases the contour accuracy of merged regions.Experimental result shows that our growing method can effectively improve the accuracy of the extraction of salient regions.