Real time statistically boundary detection

The first task for object detection and recognition is image segmentation. This process is initialized with boundary detection. In this paper we propose a boundary detection method based on standard deviation value of intensity within the window. We assume that the low value is similar colour where the computation process occur on the same object. While the high value, the computation process will occur on the object's boundary or edge. In this research we compare to other methods to evaluate the performance of proposed method. The experimental result, by using window size of 3×3 pixels, our method's computation time is 66.54 ms and using window size of 5×5 pixels, the computation time is 67.39 ms. Video rate is 15 fps. Processing with the proposed method yields less noise compared to other method, both for colour and grayscale processing.