Performance Evaluation of Pansharpening Methods on GPU for RASAT Images

Acikgoz I. S., Teke M., KUTBAY U., HARDALAÇ F.

7th International Conference on Recent Advances in Space Technologies (RAST), İstanbul, Turkey, 16 - 19 June 2015, pp.283-288 identifier

  • Publication Type: Conference Paper / Full Text
  • City: İstanbul
  • Country: Turkey
  • Page Numbers: pp.283-288
  • Keywords: RASAT, image processing, pansharpening, GPU, CUDA, FUSION
  • Middle East Technical University Affiliated: No


Turkey is among the countries which could develop earth observation satellites. RASAT and Gokturk-2 satellites are still operational and continuously acquire images of the Earth. Their images are processed before sharing with end users. Pansharpening, at which high resolution pan and low resolution multi-spectral images are fused, is an important step in image processing chain. As the resolution and number of images increase, pansharpening of satellite images take considerable amount of time. Multithread programming and General Purpose GPU (GPGPU) programming implementation improve performance of image processing applications, where most operations carried out on individual pixels. In this paper, we compared pansharpening applications and their CPU and GPU implementations for RASAT images. GPU implementations of pansharpening algorithms provides 20-25 times speed-up compared to CPU implementations.