A CLEAN Algorithm Based on Multi-Scale Bandpass Filter and Its GPU Implemention

China's new generation radio telescope-The Mingantu Ultrawide SpEctral Radioheliograph (MUSER) generates high-quality radio images with high temporal, high spatial, and high spectral resolutions in the frequency range of 400MHz-15GHz. The MUSER will provide a new observational window on researches such as solar activities and solar electron accelerations. High-performance and high-quality imaging form a significant part in aspect of MUSER's massive data processing requirements. Referring to the data processing method of the MEUDON observatory in France, we give a detailed analysis of multi-scale CLAEN. Appropriate parameters are given for MUSER CLEAN and the implementation based on GPU is presented. The experimental results show that the multi-scale CLAEN is nearly three times faster than the previously realized Högbom CLEAN, which will improve the performance of the whole data processing system effectively.