Parallel Implementation of Computer Vision Algorithms for Object Detection

U ovom radu prikazana je paralelna implementacija algoritma za detekciju objekata, točnije algoritam histogram usmjerenih gradijenata. Prilikom implementacije korišteni su principi paralelnog programiranja uz vrlo popularni MPI standard. Detaljno je objašnjen cjelokupni proces pretvaranja slijednog algoritma u paralelni te je prikazana cjelokupna ideja programske izvedbe u jeziku C++. Na koncu su izneseni i grafički prikazani rezultati testiranja sustava nad različitim brojem procesa i dimenzijama testnih slika. Također je provedena usporedba s postojećom OpenCV implementacijom HOG algoritma.This work presents a parallel implementation of object detection algorithm, more precisely histogram of oriented gradients. During the implementation of algorithm there were used many principles of parallel programming based on MPI standard. The whole process of transforming serial algorithm to parallel is explained very detailed and it is presented with implementation in programming language C++. Testing was carried out over various picture dimensions and number of processes. In the end results were provided and explained graphically and compared with OpenCV implementation of HOG algorithm