Information Technology Reference
In-Depth Information
6Con lu ion
In this chapter, we presented an extension of steering kernel regression for video
upscaling. Our proposed algorithm is capable of spatial upscaling with resolution
enhancement, temporal frame interpolation, noise reduction, as well as sharpening.
In the proposed algorithm, we construct 3-D kernels based on local motion vec-
tors, unlike our previous work [11, 13]. The algorithm includes motion estimation,
but doesn't use explicit motion compensation. Instead, the spatio-temporal kernel
is oriented along the local motion trajectory, and subsequent kernel regression acts
directly on the pixel data. In order to avoid introducing artifacts due to motion esti-
mation errors, we examine the motion vectors for their reliability. We apply a tem-
poral weighting scheme, which allows us to suppress data from neighboring frames
in the case of a motion error. Also, we reduce the computational cost of MASK
by using a block-based motion model, using a quantized set of local orientations,
and adapting the regression order. The adaptive regression order technique not only
reduces the computational cost, but also provides sharpening while avoiding noise
amplification.
We have presented several video upscaling examples showing that the MASK
approach recovers resolution, suppresses noise and compression artifacts, and is ca-
pable of temporal frame interpolation with very few artifacts. The visual quality of
the upscaled video is comparable to that of other state-of-the-art multi-frame up-
scaling methods, such as the Non-Local-Means based super-resolution method [11]
and 3-D ISKR [13]. However, the computational complexity of MASK in terms of
processing and memory requirements is significantly lower than these alternative
methods. In order to improve the visual quality of MASK further, it may be nec-
essary to include more accurate motion estimation, for example by using smaller
block sizes (currently 8
×
8), or extending the motion model, e.g. to an affine model.
References
1. Mitchell, D.P., Netravali, A.N.: Reconstruction Filters in Computer Graphics. Computer
Graphics 22(4), 221-228 (1988)
2. Park, S.C., Park, M.K., Kang, M.G.: Super-Resolution Image Reconstruction: A Techni-
cal Overview. IEEE Signal Processing Magazine 20(3), 21-36 (2003)
3. Farsiu, S., Robinson, D., Elad, M., Milanfar, P.: Advances and Challenges in Super-
Resolution. International Journal of Imaging Systems and Technology, Special Issue on
High Resolution Image Reconstruction (invited paper) 14(2), 47-57 (2004)
4. Narayanan, B., Hardie, R.C., Barner, K.E., Shao, M.: A Computationally Efficient Super-
Resolution Algorithm for Video Processing Using Partition Filters. IEEE Transactions on
Circuits and Systems for Video Technology 17(5), 621-634 (2007)
5. Gersho, A., Gray, R.M.: Vector Quantization and Signal Compression. Kluwer Academic
Publishers, Boston (1992)
6. Hardie, R.: A Fast Image Super-Resolution Algorithm Using an Adaptive Wiener Filter.
IEEE Transactions on Image Processing 16(12), 2953-2964 (2007)
