Abstract:
This thesis proposes a Discrete Wavelet Transform (DWT) based Block Matching
Binary Motion Estimation (BME) algorithm. By using the DWT analysis, the size of
data involved in the algorithm is enormously reduced. On the process of extracting the
motion vector (MV), the block matching is performed only on one of the sub-frames of
the DWT of the original frame. Motion vector is then found by means of minimum Sum
of Absolute Differences (SAD). The proposed algorithm skips unnecessary steps of
computations so it decreases the computational load with successive elimination of noncandidate
blocks in the search window.
The proposed technique outperforms conventional full search block matching method,
which is described in MPEG-4 verification mode, in terms of both peak signal to noise
ratio and computational complexity. It is shown that the proposed algorithm is simple
and efficient while keeping the same regularity and good performance.