Existing methods to up-sample depth maps suffer from known deficiencies such as texture copying and edge blurring. All attempts to correct these deficiencies introduce filter methods that are significantly dependent on correctly selecting the window size for the filter.
With depth maps an integral component of 3D video, commercial applications of an improved method to determine depth. UCSD inventors have come up with an algorithm that reduces the typical depth map deficiencies, while also avoiding introducing a restrictive dependence on the selection of the filter size.
A prototype has been developed and tested with the Middlebury datasets (http://vision.middlebury.edu/stereo/) to great success.
Country | Type | Number | Dated | Case |
United States Of America | Issued Patent | 9,349,169 | 05/24/2016 | 2012-277 |
video processing