Pre-Processing For The Motion Detection Of Sequential Images

This thesis makes a discussion on auto-detection of motion displacement. Change detection thus becomes the key issue. Conventional simple image subtraction detects all difference regardless of their source and is problematic to use. So, it is necessary to isolate the local difference due to the motion from the global difference due to the illumination variations. Scattergram is used to resolve this problem. Then the noise is removed by means of the variance that is estimated directly from the median of the noise. Finally, an auto-detection is made for the motion displacement from the difference image undergone the above pre-processings.