文章基本信息

标题：THE PROPOSED IMAGE SEGMENTATION METHOD BASED ON ADAPTIVE K-MEANS ALGORITHM
本地全文：下载
作者：INTEDHAR SHAKIR NASIR
期刊名称：Journal of Theoretical and Applied Information Technology
印刷版ISSN：1992-8645
电子版ISSN：1817-3195
出版年度：2018
卷号：96
期号：21
出版社：Journal of Theoretical and Applied
摘要：Image Segmentation is a significant process in image analysis, which refer to partition an image into coherent regions called (segments). Image segmentation is a mostly useful task in computer vision applications, which used commonly in several applications like image compression, object tracking, object detection, and so on. Current image segmentation techniques, either required prior information about the number of desired parts or segment the image based on certain criteria like uniform texture or color. Current research works, focused on segmentation to classifying the images based on extracted objects, which help to improve retrieving process in advance search engine. The difficulty in segmentation process is how to known the number of coherence regions in the given image. No one can achieve this process except the human mind, and the human only can decided what the interesting or unusual objects in the image. However, this paper suggested a new approach by combine two famous segmentation approaches, which are, region growing based method and clustering based method. The first approach aims to segment the image through sequence of image transformation procedures, then the connected component typically the objects regions in that image. Hence, by count these regions in the image; we can estimate the number of objects in the given image. By knowing the estimated number for the objects in the given image, second approach consider this value in for evaluation process. K-Means++ typically implemented in initial step to initialize the seeds when applying standard K-Means algorithm. After the initializing step, standard K-Means algorithm used by consider the pixels� color properties at CIE color space. Both algorithms takes into consideration the SSE as a base metric to estimate the number of clusters (objects) in the image. This approach is very useful to understanding images and gives a good perception about it. Finally, the proposed system has tested and evaluated using Barkley dataset, and the experimental results have analyzed using accuracy measure. The evaluation metrics and experimental results shows that the proposed system has achieved better accuracy in order to segment the given images when compared with traditional segmentation methods.
关键词：Image Segmentation; Images Analysis; K-Means++; Region Growing; SSE; K-Means