A Hybrid Wavelet-based Motion Estimation Algorithm for Mobile Devices

In recent years there is a surge in demand for multimedia content like video-on-demand on personal digital assistance (PDAs), mobile telephones, and other mobile devices. However, these mobile devices have several constraints that is, limited processing, low display resolution, limited storage capacity, and relatively limited communication speed. While technological advances will reduce some of these constraints, mobile devices are likely to remain significantly less capable than their desktop counter parts. Therefore, research in video compression is necessary to reduce the storage memory.

In this research, a hybrid algorithm, utilizing the Dual Tree Complex Wavelet Transform (DTCWT) and the Adaptive Root Pattern Search (ARPS) block is used to perform the motion estimation. This new proposed algorithm first transform each video sequences with DTCWT. Real low pass sub-band filters are used to carry out the transform. The frame n of the video sequence is used as a reference input and the frame n+2 is used to find the motion vector. Next, the ARPS block search algorithm is carried out and this followed by an inverse DTCWT. The motion compensation is then carried out on each inversed frame n and motion vector.

The results show that a good compromise was achieved between computational complexity and image quality for mobile device which has limited memory without depriving its quality. The proposed algorithm also takes less memory usage compared to the DCT-based algorithm.