CAJ | 학술논문

为提高纺织CAD技术,依据半监督聚类理论,提出一种以最小错误率贝叶斯决策为准则的半监督聚类的织物图像分割算法.这种算法利用有限的人工信息,即在织物图像上点击有限的几个点以标识相应区域之间的关系,从而得到满足用户给定限制的织物图像分割结果.用该算法首先对织物图像进行量化转换处理,而后在贝叶斯模式识别中集成先验的分割信息进行色彩聚类.实验结果表明,该算法用于织物图像分割是可行的.
위제고방직CAD기술,의거반감독취류이론,제출일충이최소착오솔패협사결책위준칙적반감독취류적직물도상분할산법.저충산법이용유한적인공신식,즉재직물도상상점격유한적궤개점이표식상응구역지간적관계,종이득도만족용호급정한제적직물도상분할결과.용해산법수선대직물도상진행양화전환처리,이후재패협사모식식별중집성선험적분할신식진행색채취류.실험결과표명,해산법용우직물도상분할시가행적.
To improve the technology of the textile CAD, we proposed a new way of textile image segmentation by the minimum error Bayes decision theory based on the semi-supervised clustering. The algorithm makes use of limited human assistance, i.e. indicating the relationship of some different regions by clicking some limited points on the textile image using a mouse, to get the final accurate results of segmentation which satisfies the requirement of customers. The algorithm quantifies the textile image. Then it clusters colors in Bayes decision with prior segmentation information. The results demonstrated that this algorithm is a feasible way for textile image segmentation.