This repository maintains for the source code of the work

• Chang, P.-L., Stoyanov, D., Davison, A.: Real-time dense stereo reconstruction using convex optimisation with a cost-volume for image-guided robotic surgery. MICCAI 2013.

If you use the code, please cite the paper. Thank you.

Compile the CUDA MEX-Files in the directory gpu, see: mex cuda setup

In MacOS for example, you have to:

1. Copy mexopts.sh from the default path to gpu
2. Find the CUDA SDK header directory /usr/local/cuda/samples/common/inc

In OSX MATLAB R2013a, the mexopts.sh is in

/Applications/MATLAB_R2013a.app/toolbox/distcomp/gpu/extern/src/mex/maci64

Depends on your CUDA SDK you may have to remove all

-gencode=arch=compute_13,code=sm_13 in mexopts.sh.

To compile:

mex -v -O -largeArrayDims -I/usr/local/cuda/samples/common/inc HuberL1CVPrecond_mex.cu

Note that using -largeArrayDims is neccessary.