Parallelization of High Dynamic Range Image Creation on Multi-core Processor Architectures

The emergence of multi-core processor architectures and of diverse parallel computing paradigms has permeated into the area of mainstream computing. In this paper we present various parallelization approaches to High Dynamic Range image creation, a rising technology employed in the field of imaging manipulation and processing. OpenMP and Pthreads implementation details are provided, and the performance and load-balancing capabilities of each approach is discussed, together with a scheme for overlapping I/O intensive regions with CPU-bound sections of the code, on three relevant datasets with resolutions spanning from 1:2 to 20 Mega Pixels.