Object Contour Extraction Based Salience Detection and Automatic Region Growing
Tingting Lan, Shanshan Gao, Jing Chi, Yuanfeng Zhou · 2016
Contraposing to the problem of manual selection of seeds and inaccurate object contour detection for the traditional region growing algorithm, in this work, a novel object contour extraction method is proposed based on salient region detection and automatic region growing, which consists of four crucial steps. Firstly we partition off an original image to be super-pixels by hexagonally arranged iterative clustering (HAIC). Secondly, we locate the salient object by super-pixel global contrast (SGC), and then determine the centroid and background color set. Thirdly, seed can be automatically selected from background color set. Finally, the object contour is extracted by post-processing: open operation and isolated region elimination. Experimental results show that the proposed method is easy to implement with low time complexity, and the salient object contour nearly fit target boundary.