A pre-filtering algorithm for reducing aliasing artifacts is proposed in this paper. Multiview coding is the new extension of H.264/AVC, and with a lot of research effort. In multiview coding environments, aliasing problems arise. The paper proposes a novel algorithm to solve this problem, and the algorithm has two steps. The first step is oversampling of video signals. The second step is a low-pass filter to remove the ghosting artifacts caused by aliasing. There are two advantages to this algorithm: the reduction of aliasing artifacts, and the bit reduction after filtering. Some experiments are done using the multiviewcoding tools and the proposed pre-filter. The experiments and subjective evaluation show that people prefer coding with the proposed pre-filter at low-bit rate. At high-bit rate, people prefer coding without the pre-filter.
The pre-filter is a low-pass (Gaussian) filter to remove the high frequency signal in the source image. It reduces the aliasing artifacts, but it also blurs image details. The paper realizes this drawback, although it does not provide a solution. A possible direction for further improvement is to consider an adaptive filtering scheme. More research on multiple filtering algorithms is also desirable. It is interesting to compare the performances of different filtering algorithms.
This paper is recommended for engineers interested in digital filter design for multiview coding. It will prove helpful for people in the area of video compression.