A new segmentation method of CR images based on discrete wavelet transform and mathematics morphology

Abstract

In this paper, we propose a segmentation method of CR(computed radiography) images with being rid of under-segmentation and over-segmentation. An under-segmentation occurs when pixels belonging to different objects are grouped into a single region. Such errors are the most dangerous because they can invalidate the whole segmentation process. The phenomenon always takes place when we segment CR images because of the principle of CR. In order to depressed under-segmentation, we enhance the CR images using DWT (discrete wavelet transform) to get more detail of CR image features. As we enhance the CR image, the over-segmentation maybe occurs. Compared with under-segmentation, the over-segmentation occurs when a single objects is subdivided by segmentation into several region. For the purpose of preventing from the over-segmentation, we present a scheme for enhanced CR images based on watershed algorithm that solves over-segmentation problem. We use marker-based watershed algorithm. Together with gradient image and marker image, watershed segmentation can make sure to partition CR image into meaningful object and avoid further segmentation of homogeneous regions. The result of the proposed algorithm are compared with those of other standard methods. Experiments have shown a better result and indeed solved under-segmentation and over-segmentation problems.