Two-bit transform for binary block motion estimation

A Survey on Motion Estimation Techniques

“…This damages the effectiveness of the MSE matching score, which essentially integrates the squared error contributions from all frequency components. 1-bit and 2-bit transforms have been proposed before, to improve the computational efficiency of the motion estimation step in video compression application [14] [15]. [16] describes a 2-bit transform for increasing the robustness of motion estimation (registration) in a multi frame enhancement application, this transform is most similar to the one proposed here [17].…”

“…Motion estimation and compensation play key roles in video coding systems due to the ability of realizing high compression rates achieved by removing temporal redundancies between successive image frames. The most popular technique for motion estimation is block matching [28]. In the block matching algorithm(BMA), the image frame is commonly divided into non overlapping rectangular blocks.…”

“…The optimal solution to BMA is the full search (FS) algorithm that exhaustively searches for the best matched block within all locations of the search window. The mean absolute difference (MAD) or mean square error (MSE) matching criteria are considered to be statistically optimal solutions to the matching process [28]. The search for motion vectors is typically carried out in a restricted local neighborhood of the corresponding block, local transformations result in more efficient and appropriate bit plane representations compared to global trans-formations.…”

Successive elimination algorithm for two-bit transform-based motion estimation

“…The techniques that exploit different matching criteria instead of the classical sum of absolute differences (SAD) such as one-bit transform (1BT), multiplication-free 1BT, and 2BT were proposed to make the faster computation of the matching criteria using Boolean exclusive-OR (XOR) operations [4,5,6]. In [4], 1BT-based ME where the reference frames and the current frames are transformed into one-bit representations by comparing the original image frame against a bandpass filtered output was proposed.…”

“…2BT-based ME was proposed to enhance the ME accuracy of 1BT-based ME algorithms [6]. In 2BT-based ME, the values of local mean μ, variance σ 2 , and the approximate standard deviation σ a are used to convert frames into two-bit representations which are calculated as [6]:…”