Minimum Rate Predictors family of encoders

Working implementations of various codecs based on Minimum Rate Predictors (MRP) resulting from my research activities at Instituto de Telecomunicações - Leiria Branch.

Codecs implemented in this code (each in a different branch):

• MRP Video [1], master branch
• 4D-MRP [2], 4DMRP branch
• DT-4D-MRP [2], 4DMRP-DUAL-QUADTREE branch
• M-MRP [3], lenslet branch
• H-MRP [4], HiMRP branch

Requirements

In the source code of each prediction mode a CMakeLists.txt file is provided. The only requirements are a working CMake and C/C++ environment.

How to use?

• Each implemented codec has a different git branch.
• After selecting the desired branch create a sub folder to run CMake, example for linux:

    mkdir build
    cd build
    cmake ../

• Each binary can present a help text when run without options, such as the following example of 4D-MRP:

    IT - Leiria: Minimum Rate Predictors Lenslet
    encmrp/decmrp version 1.1.0 (February 2019)
    usage: encmrp [options] infile outfile
    options:
        -J 2 * num  Views dimensions (in pixels) [H W]
        -K 2 * num  Dimensions of the array of views [H W]
        -L 5 * num  Prediction order (1 * Intra, 4 * Inter) [-1 0 0 0 0]
        -b num      Bit depth [8]
        -E num      Endianness: little-endian = 0, big-endian = 1. Default: little-endian
        -C str      Chroma format [GRAY]. Supported formats:
                    GRAY;
                    444;
                    422; --> Not yet implemented
                    411; --> Not yet implemented
                    420.
                    (Notice: Currently MRP is a Luma only encoder. Thus this step is used only to skip the Chromas.)
        -D num      Distance between views [1]*
        -M num      Number of predictors [-1]
        -P num      Precision of prediction coefficients (fractional bits) [6]
        -V num      Number of probability models [16]
        -A num      Accuracy of probability models [3]
        -I num      Maximum number of iterations [100]
        -m          Use MMSE predictors
        -h          Use Huffman coding
        -f          Fixed block-size for adaptive prediction
        -u          Deactivate the histogram packing procedures
        -o          Further optimization of predictors (experimental)
        -d          Create extra debug output (coefficients, partitions, etc.)
        -r str      Light field file format [SAI]. Supported formats:
                    MIA; --> Not yet implemented
                    PVS;
                    SAI.
    infile:         Input file (must be in a raw YUV format)
    outfile:        Output file

    Note: * stands for a mandatory option.enter code here

• The MRP family encoders compress each image component separately.
• Each component should be provided as a raw file, different codecs have different requirements:
  • MRP Video: single image or video sequence (the resolution height and width must be multiple of 8);
  • M-MRP: LF re-arranged as an array of SAIs (the resolution height and width must be multiple of 8);
  • 4D-MRP, DT-4D-MRP, and H-MRP: LF re-arranged either as a pseudo-video sequence of SAIs or as an array of SAIs.
• The configurations used for each codec can be found on their associated publications.

References

[1] J. M. Santos, A. F. R. Guarda, L. A. da Silva Cruz, N. M. M. Rodrigues, S. M. M. Faria, Compression of medical images using MRP with bi-directional prediction and histogram packing, in: Picture Coding Symposium (PCS), Nuremberg, Germany, 2016, pp. 1–5.

[2] J. M. Santos, L. A. Thomaz, P. A. A. Assunção, L. A. da Silva Cruz, L. Távora, S. M. M. Faria, Lossless Coding of Light Fields based on 4D Minimum Rate Predictors, in: IEEE Transactions on Image Processing, vol. 31, pp. 1708-1722, 2022.

[3] J. M. Santos, P. A. A. Assuncao, L. A. da S. Cruz, L. M. N. Tavora, R. Fonseca-Pinto, S. M. M. Faria, Lossless compression of Light Fields using multi-reference Minimum Rate Predictors, in: Data Compression Conference (DCC), Snowbird, UT, USA, 2019, pp. 408–417.

[4] J. M. Santos, L. A. Thomaz, P. A. A. Assuncao, L. A. da S. Cruz, L. M. N. Tavora, S. M. M. Faria, Hierarchical Lossless Coding of Light Fields with Improved Random Access, in: Signal Processing: Image Communication, vol. 105, pp. 116687-116687, 2022.