Determining the Number of Regions for X-ray Image Segmentation by Using Image Projection
基于投影的X光图像分割区域数量判定

Automatic detection of low intensity threat objects in the x ray image of carry on luggage requires firstly segmenting the x ray image into a number of object regions. The segmented image should contain sufficient objects, neither over segmented nor less segmented. In general, the optimal number of regions is not known in advance. The widely used statistical validity index methods can not give satisfactory estimation of the optimal number of regions because they do not consider the spatial information of the image. A novel method for determining the optimal number of image segmentation regions by using image projection is proposed. The new method uses the correlation between both the horizontal and vertical projections of consecutive segmented images to determine the optimal region number in the segmented image. Its advantage over the statistical validity index methods is that it utilizes both the statistical and spatial information of the image. Experiment results show that the proposed method produces segmented image with sufficient object information and has low computational requirement.