This allows a stereo pair of images captured with a calibrated pair of cameras to be processed automatically to derive a dense 3D model, with a 3D point calculated for every foreground pixel. This process does not require any markers, makeup, structured light or random dot projection. Synchronised video cameras are used to capture sequences of stereo pairs of images at up to 60 fps (and even up to 1,000 fps with a custom camera system), which can then be processed to recover automatically a sequence of 3D models.
Multiple stereo pairs of cameras can be used to capture multiple overlapping 3D models, which are then merged automatically into a single 3D model per frame. If colour cameras are used then a per frame colour texture map can be applied to the 3D model sequence.
Processing can be carried out entirely on the CPU or can be accelerated by processing on a GPU using NVIDIA’s CUDA. Processing can also be accelerated by distributing it over a network of processing PCs.
In addition to calculating per frame 3D models, the DI4D Processing Software also calculates a dense optical flow map for each image to the previous and subsequent image in the sequence. This dense optical flow data is used in the DI4D Tracking Software to track landmarks or the vertices of a fixed topology mesh through a sequence.
Michael Illingworth, Owner, Vine FX
Experience high fidelity facial performance captureRequest a Quote