Component Tree Based Image Processing Methods And Applications

Mathematical morphology is an important image processing method.The component tree is a further development of mathematical morphology theory and methods.Unlike traditional structural element-based methods,component trees perform image processing based on the concept of regions.The component tree is a general representation of image contents,which generates the connected components of the grayscale image through the upper thresholds and organizes the threshold sets using the structure of rooted tree.Component tree can be used to represent the hierarchical relationship between the image regions.The component tree is a good low-level image processing method,which can provide high-efficiency preprocessing methods for high-level semantic extraction of images.This thesis studies the image processing methods and applications based on component tree.The advantage of this method is that the processing is performed on the constructed component tree by graph transformations,and only the component tree structure is used from low-level to high-level processing.The main work of this paper is as follows.Aiming at the representation of gray-scale image component tree,an efficient node-oriented representation method is proposed.Node index and node position are used as essential factors to represent the component tree nodes,to solve the issues to convenient nodes traversal and processing for the gray-scale image component tree.A classification method for hand posture images is proposed.Firstly,an efficient gray image component tree construction algorithm is used to build component tree for posture images.Since the original image usually has noise,the structure of the component tree is relatively complex,and there are a large number of small-area nodes.Therefore,the component tree is simplified by using various filtering rules based on node attributes.The node corresponding to the gesture area are then located from the simplified tree,and then server useful attributes of the component tree nodes is extracted as the feature description of posture images.Finally,the machine learning method is used to train a classifier,and the classification accuracy of the classifier is verified by experiment.Experimental result shows that,the component tree filtering and simplification methods used in this paper can be used to extract the region of interest in the image.The extracted component tree node attributes can be effectively used as image feature description and used for image classification tasks.