Motion Analysis by the RHT Method

Licentiate Thesis ii Abstract Motion analysis is an important task in computer vision. In this work, a short overview of motion estimation techniques is presented and motion detection by a new Hough-like technique, the Randomized Hough Transform (RHT) is studied. The Standard Hough Transform and its applications to motion estimation are shortly introduced. The ideas of the RHT are presented and some new probabilistic Hough techniques are compared. Then, the RHT is applied to motion analysis. A new method is developed to calculate 2-D motion in a sequence of time-varying images. The method, called Motion Detection using Randomized Hough Transform (MDRHT), is based on the Randomized Hough Transform. The RHT decreases considerably the time consumption and memory requirements of the Hough Transform. The basic idea of the MDRHT is to pick randomly point pairs from two images and calculate the translation with them. The points can be e.g. edge points of the original images. This approach can avoid diiculties of standard segmentation methods, e.g. overlapping and covering, and has the advantages provided by the RHT. The method can be generalized by picking more than two points. The extended algorithm to calculate both the translation and the rotation is also represented in this work. Probabilistic mechanisms of the MDRHT are also considered and new possible extensions are introduced. Algorithms are tested with both real and synthetic images in the presence of noise and their behaviour is examined via probability formulas deened in this work. Preface The work presented in this thesis for the degree of Licentiate in Technology has been carried out at the Laboratory of Information Processing in the Depart-thank student Petri Hirvonen for his help in computer programming. The nancial support of the Academy of Finland and Lappeenranta University of Technology is gratefully acknowledged. Finally, special thanks for my ancee Virpi for her support and patience.