改进的空间约束加权模糊核聚类红外图像分割

Improved Weighed KFCM Based on Spatially Constrained for Infrared Image Segmentation

红外图像分割算法对复杂背景下的目标检测跟踪具有重要意义,提出了一种改进的基于空间约束的加权模糊核聚类红外图像分割新算法.在其中引入了红外图像像素间的空间位置约束关系和关于类别的结构信息,并定义了类别权重可靠性指数修正类别权重,不但抑制了红外图像中存在的噪声点和野值等干扰,而且可以保护红外图像中的小目标,防止被背景淹没.通过对实际红外图像的分割结果表明,该算法很大程度上减少了背景像素对目标识别的干扰,适于进行复杂背景下红外目标的准确分割.

The method of infrared image segmentation is important for detection and tracking in complex background.An improved method for infrared image segmentation based on weighted fuzzy kernel clustering using spatial relation was proposed.The pixel＇s spatial positions restriction and the structure information about class were introduced.And a reliability of class weight was defined to revise the label weight factor.The noise of infrared and wild value was suppressed.Simultaneously the small targets were protected,in order to prevent submerging by background pixel.Through the experimental on real infrared image showing the obstruction to targets recognition by background pixel was decreased largely by the proposed method.It is be fit for infrared targets segmentation accuratly in complex background.