CERV Representations


Appendix

Description of the algorithm for optimal pruning of the context tree for crack-edges: pdf

Additional Figures and Tables

Comparative results for lossless compression of the depth-map datasets: Comparisons are made with: REFERENCES
[1] S. B. Kang. MSR 3D Video Download.
[Online]. Available: MSR 3D Video Download.

[2] Middlebury.
[Online]. Available: The Middlebury repository Download.

[3] Philips.
[Online]. Available: Philips repository Download.

[4] The Piecewise-Constant Image Model (PWC), Paul J. Ausbeck Jr.
[Online]. Available: PWC.

[5] Hewlett-Packard Labs. HP labs LOCO-I/JPEG-LS home page.
[Online]. Available: LOCO-I.

[6] Xiaolin Wu. Context-based Adaptive Lossless Image Codec.
[Online]. Available: CALIC.

Comparison with other programs (filesize)

The results from five programs are compared:
CERV-HiCo, CERV-Fast, PWC, CALIC and LOCO-I.

Depth-map dataset:

Comparison with other programs (runtime)

The results from four programs are compared:
CERV-Fast, PWC, CALIC and LOCO-I.

Depth-map dataset:

Comparison between CERV profiles

The results from four CERV profiles are compared:
CERV-HiCo, CERV-Fast, CERV-2 and CERV-3

Depth-map dataset:

Additional information for the paper

The Middleburry database [3] has 27 datasets for which disparity maps were provided as ground truth, containing the 2005 Stereo dataset (Art, Books, Dolls, Laundry, Moebius, Reindeer) and the 2006 Stereo dataset (Aloe, Baby1-3, Bowling1-2, Cloth1-4, Flowerpots, Lampshade1-2, Midd1-2, Monopoly, Plastic, Rocks1-2, Wood1-2). In each dataset taken from [3] there are disparity maps taken from two views, and for three resolution levels: full-size (width: 1330..1390, height: 1110), half-size (width: 665..695, height: 555), and third-size (width: 443..463, height: 370).

The next most frequently occurring non-deterministic contexts, following the first three contexts: .pdf
Last update: June 14, 2013
WebDesign: Ionut Schiopu