FAST AND ACCURATE FEATURE DETECTION AND TRIANGULATION USING TOTAL VARIATION FILTERING OF BIOLOGICAL IMAGES

摘要: We consider here the problem of detecting and modeling essential features present in a biological image construction compact representation for them which is suitable numerical computation. The solution we propose employs variational energy minimization formulation to extract noise texture, producing clean containing geometric interest. Such decomposition reduce complexity further processing. are particularly motivated by registration where goal align matching pair images. A combination algorithms from combinatorial optimization computational geometry render fast solutions at interactive or near rates. demonstrate our technique microscopy able, example, process large, 2048times2048 pixels, histology mouse brain images under minute creating faithful sparse triangulation model it having only 1.8% its original pixel count. Models 512times512 typically generated less than 5 seconds with similar reduced vertex These results suggest relevance approach biomedical