An Improvement of Motion Detection Using Hough Transform: A Multi-Resolution Approach

Hough Transform (HT) has been widely used in computer vision. One application is the use of HT to find motion parameters (dx and dy), of moving objects in motion analysis. The edge points in two image frames are compared to find their correspondents. The displacement of the corresponding points is calculated and the accumulator in the displacement space accumulates the number of displacement hits. This traditional method is highly time-consumed. This paper introduces a technique to reduce computational requirements in motion detection using HT. With this technique, a pair of low-resolution images is created from the high-resolution ones and HT is applied to the low-resolution image pair to find the object displacement. The result from this process is then used as a guideline to specify the regions needed for motion detection using HT in the highresolution image pair. Tests with synthetic and real-world images verify that the proposed method is more effective and faster than direct motion detection using HT especially when the noise pixels in the images increased.