THE POSSIBILITY OF PARALLELIZATION OF IMAGE PROCESSING USING OPENCV AND OPENMP

UDC 681.3
doi:10.26102/2310-6018/2019.25.2.011

V.V. Bernikov, A.P. Preobrazhenskiy, O.N. Choporov


The relevance of the study is due to the need to solve problems related to image processing in various technical applications. Several approaches are considered: on the basis of the usual access to image pixels, when in fact all values of the array are bypassed in turn, access to pixels is carried out using arithmetic operations on pointers, pixels are located within a single continuous memory block in a sequential manner, and, on the basis of the proposed approach associated with parallelization of calculations, using multithreading. On the basis of empirical studies, the possibility of accelerating the calculations based on the proposed method several times was shown. A block binarization algorithm is considered when binary blocks form a complete binary image. Within this algorithm parallelization of calculations is carried out. When implementing the algorithm, the C++ language and OpenCV and OpenMP libraries were used. On the basis of empirical studies in tables and graphs, it is shown that due to parallelization, even at full load of the kernel, the image processing time was reduced by almost 2 times, which confirms the possibility of using the proposed algorithm.

Keywords: :OpenMP, OpenCV, parallel computing, image processing

Full text:
BernikovSoavtori_2_19_1.pdf