A Color-Texture Segmentation Method to Extract Tree Image in Complex Scene

In order to provide basic data for tree image recognition and feature extraction, this paper studies the tree image segmentation in complex background. Based on the visual characteristics differences of the tree and the surrounding objects, the trees from different backgrounds are separated into a single set of tree image pixels. This paper proposes a segmentation method to extract object based on color and texture features of color tree images. Firstly, reduce noise of the color tree image by the use of anisotropic filter. Then, we select Lab color space as the space for image segmentation. And the color image of RGB space is transformed into Lab space. Next, due to the negative end of a-channel reflects the color feature of trees, the L, a, and b channels are split. Green is the main feature of tree images, so segmentation by two-dimension OTSU of automatic threshold in a-channel. Based on the color segmentation result, and the texture differences between the background image and the object tree, we extract the object tree by the gray level co-occurrence matrix for texture segmentation. Finally, the segmentation result is corrected by mathematical morphology methods. This method not only segmentation speed is faster, and without human participation, but also the segmentation result is ideal when there are not green plants which are so close to the object tree in the tree image background.