Implementations of Parallel Error Diffusion Optimized for GPUs (Preliminary version)

Abstract

Error diffusion is a well-known algorithm used for converting a gray-scale image into a binary image. The main contribution of this paper is to present our new error diffusion algorithm optimized for CUDA-enabled GPUs. Our algorithm partitions the gray-scale image into parallelogram blocks and performs error diffusion operation for them in parallel. We have implemented several error diffusion algorithms on GeForce GTX 780 Ti. The experimental results show that our algorithm runs faster than any other error diffusion algorithms for any input matrices. Also, it runs 44 times faster than the sequential error diffusion algorithm using a single CPU.