A Massive Parallel Computing Algorithm Based on Gpu/Cuda for Processing Large-Scale Multifractal Images

This paper presents a massive parallel computing algorithm implemented on GPU (NVIDIA CUDA) architecture for the real-time computation of multifractal parameters in high-resolution digital images. The conventional sequential computational complexity of O(NlogN) has been optimized through parallel box-partitioning and dynamic shared memory allocation within CUDA threads. Experimental results demonstrate that the proposed parallel algorithm reduces processing latency by 6.0 to 9.5 times compared to CPU-based execution.