Connected Filtering and Segmentation Using Component Trees

摘要

This paper formalizes the notion of nonflat gray-level connected filters and proposes efficient algorithms for their implementation in a supplementary web page. The component tree is proposed as an efficient and accessible data structure used to implement these filters. The primary nonflat component filter advocated in this paper is based on the concept of an attribute signature. The attribute signature captures both attribute and linking information between components in a gray-level image and can be a powerful means of discriminating desired image features. One of the key benefits of the approach is that the image features to be filtered undergo the maximum amount of filtering that is possible without altering the rest of the image at all. As a consequence, an image segmentation can be obtained simply by finding those pixels within the image that have been changed by the filter. We present an application of nonflat component filters to the segmentation of wood micrographs.