An Improved Approach to Background Removal Using Haar-Based Preprocessing for Phase Features

摘要

The paper proposes moving object detection in Haar preprocessed images by removing background initialization constraint like a change in illumination. The fundamental objective of the proposed method is to increase the processing speed. In this work, background subtraction algorithm using phase as a feature is used to model background and distance transform is used for foreground extraction. Operation is performed with and without Haar compressed images. Experiment on real-time video streams with Haar-based phase feature implementation shows that the speedup in object detection is achieved compared to baseline method. Performance improvement of about 79% is observed in processing speed of frames with only 25% additional frame memory.