Efficient Dense Stereo Matching using CUDA
Ke Zhu, Pablo d’Angelo, Matthias Butenuth · mediaTUM – the media and publications repository of the Technical University Munich (Technical University Munich) · 2012
The Semi-Global Matching method approximates a global, 2D smoothness constraint by combining many 1D constraints from different aggregation The values in the image are partitioned lineby-line and tiled into the shared memory to reduce the memory accesses on global memory and increase the data utilization rate, because each pixel from the right image can be used (disparity range 1) times. Each thread in a block answers to a C st tble