Multiscale edge detection for medical image enhancement

A general framework is presented in this paper for edge detection and enhancement of medical images. The method is based on multiscale analysis using filter banks, and it is adaptive to a large number of features. Initially, an optimal one-scale filter is designed for the required detection. This one-scale filter is further extended to a set of multiscale filters, which in turn are used in designing the filter bank that would provide the desired multiscale responses. Subsequently, the scale space information is optimally combined in a maximum-a-posteriori (MAP) classifier, whose design depends on the desired feature and the resulting filter bank. The method is robust to noisy conditions which are common to medical images in angiography, echocardiography, blood vessels, and others based on ultrasonic imaging, X-ray, and tomography.