RAHT (Region-adaptive Hierarchical Transform)

This code implements an improved version of the RAHT-rlgr coder and decoder as described by Queiroz and Chou [1].

[1] R. L. de Queiroz and P. A. Chou, “Compression of 3d point clouds using a region-adaptive hierarchical transform,” IEEE Transactions on Image Processing, vol. 25, no. 8, pp. 3947–3956, August 2016.

Digital Image and Video Processing Group
University of Brasília
Department of Computer Science
Department of Electrical Engineering

After compilation, two executables are generated:

RAHT_cod: Region Adaptive Hierarchical Transform coder (RAHT-rlgr)
RAHT_dec: Region Adaptive Hierarchical Transform decoder (RAHT-rlgr)

Their usage, how to compile the executable and testing is described below

Usage

RAHT_cod (Region Adaptive Hierarchical Transform coder)

Usage	Description
`RAHT_cod()`	prints help information
`RAHT_cod(V, C, depth, filename, Qstep)`	transforms the coefficients and write the coded data to the file given by "filename" with the given quantization step.

INPUTS	Description
`V`	(Nx3 double matrix) vertices of each voxel in the order (X, Y, Z). The values should be non negative integers
`C`	(Nx3 double matrix) colors associated to each voxel. This input is transparente to RGB, YUV, YCbCr or any other color space to be used as long as it is a triplet. It is also transparent to the color range: colors can be in the range 0--1, 0--255, 0--65535 or any other range
`depth`	(1x1 double value) the depth of the octree to be used for the transform. It needs to be an integer value grater than 0
`filename`	(string of chars) file name
`Qstep`	(1x1 double value) quantization step. It is dependent of the color range

RAHT_dec (Region Adaptive Hierarchical Transform decoder)

Usage	Description
`RAHT_dec()`	prints help information
`C = RAHT_dec(V, filename)`	decodes the data from the given file

INPUTS	Description
`V`	(Nx3 double matrix) vertices of each voxel in the order (X, Y, Z). The values should be non negative integers
OUTPUT
`C`	(Nx3) colors associated to each voxel

Compilation

There are two ways to compile the code, depending on your MATLAB version

Newer versions

mex( ...
    'd_main.cpp', ...
    'haar3D.cpp', ...
    'fixedpoint.cpp', ...
    './file/file.cpp', ...
    '-I./file', ...
    '-I.', ...
    '-output', ...
    'RAHT_cod', ...
    'COMPFLAGS="$COMPFLAGS /EHsc"')

mex( ...
    'i_main.cpp', ...
    'haar3D.cpp', ...
    'fixedpoint.cpp', ...
    './file/file.cpp', ...
    '-I./file', ...
    '-I.', ...
    '-output', ...
    'RAHT_dec', ...
    'COMPFLAGS="$COMPFLAGS /EHsc"')

Older versions

mex( ...
    'd_main.cpp', ...
    'haar3D.cpp', ...
    'fixedpoint.cpp', ...
    'sqrtif.cpp', ...
    './file/file.cpp', ...
    '-I./file', ...
    '-I.', ...
    '-o', ...
    'RAHT_cod', ...
    'COMPFLAGS="$COMPFLAGS /EHsc"')

mex( ...
    'i_main.cpp', ...
    'haar3D.cpp', ...
    'fixedpoint.cpp', ...
    'sqrtif.cpp', ...
    './file/file.cpp', ...
    '-I./file', ...
    '-I.', ...
    '-o', ...
    'RAHT_dec', ...
    'COMPFLAGS="$COMPFLAGS /EHsc"')

FOR TESTING, ON MATLAB EXECUTE:

This test will compute the transformed coefficients that will be writen to file 'test.raht', decode the colors matrix from file and display the resulting Root Mean Squared Error.

load test.mat
RAHT_cod(V, C, 9, 'test.raht', 10);
C2 = RAHT_dec(V, 'test.raht');
fprintf('RMSE: %g\n', sqrt(mean((C2(:)-C(:)).^2)))
s = dir('test.raht');
fprintf('Bytes: %d\n', s.bytes);

Expected result:

RMSE: 2.95404 Bytes: 986760

zmy1871 / RAHT