Superpixel-based Image Segmentation and Region Merging

• Main Authors: WANG,PIN-WEN, 王品文
• Other Authors: HSU,WEI-YEN
• Format: Others
• Language: zh-TW
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/tkqdsb

碩士 === 國立中正大學 === 資訊管理系研究所 === 105 === In the field of computer vision and image processing, image segmentation occupies a very important position. Image segmentation technology is constantly being put forward, however, the current image cutting still has many difficulties, and now some methods can be applied to the color image segmentation, most of the thinking only the image from one-dimensional space expansion to three-dimensional color space, and did not discuss other relevant information provided in color data. Therefore, the color space for image segmentation is also a very worthy of one of the topics of in-depth study. The purpose of image segmentation is to be able to find areas of interest from the image, or a meaningful area. Superpixel can achieve redundant information, and reduce the complexity of follow-up processing tasks, has been the growing concern of researchers at home and abroad. This study presents a segmentation approach to the SLIC superpixel approach and the sub-regions with the smallest of the eigenvalues of the H, S, V, R, G and B color characteristics combined with the texture are combined with the sub-regions, and the background is complex and the background is complex. Object and background difference of low type of color image for regional segmentation. According to the experimental results, this study suggests that the way to successfully segment the complex objects in complex background images. Finally, the results of this study and application of the results of the discussion and future prospects.