Performance evaluation of image smoothing on CPU and GPU using multithreading — An experimental approach in high performance computing

The processing techniques and time consumed during the execution of a task in CPU and GPU vary depending on the technology of the architecture and their configuration. This paper compares the time consumed and the efficiency of CPU and GPU with different architectures and configurations to apply spatial smoothing filters on the images of a fixed spatial resolution. The processing speed was increased by 20 to 40 times, when GPU was used rather than CPU. Moreover, the processing time decreased further when filter specific configurations were used.