Abstract
Implements an approximate nearest neighbor algorithm for many-dimensional points, called Locality Sensitive Hashing in C, then optimizes it using OpenACC. One of the goals of the thesis is to see how well OpenACC is able to parallelize such a program. The other goal is to see how well the LSH algorithm, when optimized by OpenACC, is able to solve the nearest neighbor problem for SIFT feature descriptors, with the hope that this can be used for depth estimation in the AliceVision project.