Gaussian Filtering for Computational Photography
Published in
1 min readOct 7, 2017
- Ways to do approximation algorithms
Because the Gauss transform can express a rich and useful family of image processing algorithms, but its use has been hindered by its computational complexity; instead, approximate algorithms plays vital role in the application of GF into real use.
The first data structure is permutohedral lattice, which has O(d²n) time and space complexity; The second one is the Gaussian kd-tree, which has O(dnlogn) space and time complexity. For the dimensions are between 3–12, PL > KDT, when the d>12, then we go for the KDT.
