Application Research Of Rough Set Theory In Color Image Segmentation

Image segmentation is a key step of image analysis and recognition,and is an indispensable process of image processing to image understanding.As an important image information,Colors play a key role in color image segmentation.Nowadays,most color image segmentation algorithms use the gray image segmentation algorithm directly without considering color information.As a kind of data analysis theory,rough set is used to deal with incomplete and uncertain information,which doesn't require any prior knowledge and has a strong ability of acquiring information.Due to the complexity of image itself,incomplete and uncertain information inevitably appears in each image processing.Therefore,more and more scholars apply rough set to color image processing.The main content of this paper is application research of rough set in color image segmentation,including the following three aspects:(1)The paper summarizes the classical color image segmentation algorithms,analyzes the advantages and disadvantages of those algorithms,and makes a summary of the color spaces.It also introduces the basic concepts of rough set,explores the application of rough set to color image segmentation.(2)A novel image roughness histogram is defined,and a color image segmentation algorithm is proposed based on hierarchical idea and rough set in the HSI space.First,in intensity component,the rough segmentation is implemented based on image roughness histogram.Second,in hue component,the fine segmentation is implemented on the basis of the rough segmentation results.Finally,the region merging is completed in the RGB space.Experimental results show that the proposed algorithm is effective,the detailed small target areas are clearly separated,and the computing rate is also improved.(3)A color image segmentation algorithm is proposed based on rough fuzzy set.First of all,the image is divided into a number of equal size windows in hue component,the new roughness histogram is defined based on the rough fuzzy set and is applied to image initial segmentation.Secondly,in intensity component,image fine segmentation is completed based on the results of initial segmentation.Finally,the over-segmentation regions are merged in accordance with certain criterion.Experimental results show that the presented method has better performance.