Image Segmentation Using HIS Color Space and Considering Pixel Block Properties

Abstract

In this paper, we propose a novel approach to segment image using HIS color space and considering pixel block properties. HIS color space is used to express the pixel color information. Considering the hue value will be very sensitive when the saturation and intensity values are low, the color purity is used as the weight on the hue value. This method considers not only the pixel properties, such as color and position, but also the pixel block properties. An image is divided into blocks. For each block, the mean and variance values are calculated. The variance value can be seen as the pixel block property that expresses the variation of pixels in the block. Based on the block mean, the image is segmented with the region growing approach. The pixels in the small objects obtained from the image segmentation will be merged into the large object by the decision of vector distance together with the pixel position information. The experimental results confirm this approach is useful and suitable for many kinds of images.