講演抄録/キーワード |
講演名 |
2014-06-24 10:05
Improved Interactive Medical Image Segmentation using Graph Cut and Superpixels ○Titinunt Kitrungrotsakul・Chunhua Dong・Xian-Hua Han・Yen-Wei Chen(Ritsumeikan Univ.) MI2014-25 |
抄録 |
(和) |
Interactive image segmentation is a useful method for selecting object of interest in image. The variations of intensity and shape in medical images (organs) limits their ability to precisely localize object boundaries, computation time of segmentation and therefore lack of accuracy in the segmentation object. The popular interactive segmentation method is Graph Cut. The computation time of each cut is a key to make interactive image segmentation useful in real application usage. The generally of medical images are larger than 2D image. The lack of computation time will be occur if we try to apply segment out the object in the medical images using only Graph Cut. This paper presents a method for combining Graph Cut with SLIC (Simple Linear Iterative Clustering) to adapt to medical image. To be precise, our method is initialized by design superpixels with SLIC super pixels. With SLIC superpixels, we can increasing the accuracy and also boost up computation time of Graph Cut. The experiments show segmentation results with our method is significantly better than only using Graph Cut in term of accuracy and computation time. |
(英) |
Interactive image segmentation is a useful method for selecting object of interest in image. The variations of intensity and shape in medical images (organs) limits their ability to precisely localize object boundaries, computation time of segmentation and therefore lack of accuracy in the segmentation object. The popular interactive segmentation method is Graph Cut. The computation time of each cut is a key to make interactive image segmentation useful in real application usage. The generally of medical images are larger than 2D image. The lack of computation time will be occur if we try to apply segment out the object in the medical images using only Graph Cut. This paper presents a method for combining Graph Cut with SLIC (Simple Linear Iterative Clustering) to adapt to medical image. To be precise, our method is initialized by design superpixels with SLIC super pixels. With SLIC superpixels, we can increasing the accuracy and also boost up computation time of Graph Cut. The experiments show segmentation results with our method is significantly better than only using Graph Cut in term of accuracy and computation time. |
キーワード |
(和) |
segmentation / superpixel / Graph Cut / SLIC / interactive segmentation / / / |
(英) |
segmentation / superpixel / Graph Cut / SLIC / interactive segmentation / / / |
文献情報 |
信学技報, vol. 114, no. 103, MI2014-25, pp. 17-20, 2014年6月. |
資料番号 |
MI2014-25 |
発行日 |
2014-06-17 (MI) |
ISSN |
Print edition: ISSN 0913-5685 Online edition: ISSN 2432-6380 |
著作権に ついて |
技術研究報告に掲載された論文の著作権は電子情報通信学会に帰属します.(許諾番号:10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
PDFダウンロード |
MI2014-25 |