A texture-based approach to chest radiography segmentation

Segmentation is an important pre-processing step for many inspections or detections in image analysis. This paper presents a new approach where we integrate statistical features in the segmentation process of chest radiography, which is the first step towards a computer-aided diagnosis system. In this proposal, instead of using the intensity of a given pixel, a statistical feature vector is extracted from a sliding window centred on the pixel, and then a multilevel threshold segmentation algorithm is applied. Experimental results are given to demonstrate the improved performance of the proposed method.