Integral-map Based Fast and Robust Screening Method for Endoscope image
Jia Gu wang lei · 2009
Texture-based analysis and classification is of vital importance to the early det ection of colorectal cancer precursors: flat lesion. However, the traditional way of its computation is very time consuming, which limits the application of using automated CAD (computer aided detection) algorithms in the clinical screening practices, whose one of the major concerns is high detection speed. Therefore in this paper, we present a fast yet robust algorithm for texture classification using discriminative learning models. By adopting the idea of integral-map, the feature extraction and description stages can be dramatically accelerated, thus efficient detection and screening of flat lesions in endoscope images can be ensured. Experimental results show the effectiveness and stability of our approach.