A Reduction Method of Over-Segmented Regions at Image Segmentation based on Homogeneity Threshold

摘要: In this paper, we propose a novel method to solve the problem of excessive segmentation out segmenting regions from an image using Homogeneity Threshold(). The algorithm previous based on was carried region growth by only center pixel selected window. Therefore it caused resulting in segmented regions. However, before carrying growth, proposed first all finds whether window is homogeneity or not. Subsequently, if carries total pixels But not homogeneity, So, can reduce remarkably number . order show validity method, multiple experiments compare with same environment and conditions. As results, above 40% doesn`t make any difference quality visual when method. Especially, united descending size experimentation even though has more than 1,000, can`t recognize what means. less 10, is. For these reason, expect that will be utilized various fields, such as extraction objects, retrieval informations image, research for anatomy, biology, visualization, animation so on.