Non-invasive plaque characterization of the carotid wall is crucial for the early assessment of pathology, as well as for the monitoring of the progression of a degenerative phenomenon. Specifically, in clinical practice the carotid wall status is assessed by means of B-Mode ultrasound scans. We recently implemented an algorithm for the segmentation of the common tract of the carotid wall using ultrasound relative to healthy subjects. This paper presents a superior strategy for plaque characterization, which accurately determines both echolucent-type II and echogenic plaques in pathologic subjects. We preserve both user-independence and pixel fuzziness in our approach, thereby designing an accurate intima-media thickness (IMT). Our database consists of 20 subjects comprising of normal, stable (echogenic) and unstable (echolucent) plaques. In this database of 45 images, we demonstrate our performance with respect to the gold standard tracings to an accuracy determined as normalized error to be about 8%. The results are very promising and this algorithm is being integrated into clinical setup for automatic pathologic carotid wall analysis.