Emulation of LISS III images for high temporal resolution at larger swath
- National Remote Sensing Centre, Indian Space Research Organization, Hyderabad 50037, India
Keywords: Spatio-temporal Data Fusion, Swath Expansion, Single-Image-Super resolution, AWiFS and LISS III sensors Data
Abstract. Space borne sensors have limited capability to acquire images at high spatial and high temporal resolutions with larger swath simultaneously. In this paper, we propose alternatives to overcome this limitation by emulating such images at ground data processing system. Resourcesat-2, one of the Indian Space Research Organization's (ISRO) mission carries Linear Imaging Self-Scanners (LISS III and LISS-IV) and an Advanced Wide-Field Sensor (AWiFS) onboard. The spatial and temporal resolutions of LISS III are 23.5 m and 24 days, and those of AWiFS are 56 m and 5 days, respectively. The 141 km swath of LISS III data is overlapped with the 740 km swath of AWiFS data at centre portion in simultaneous acquisition. Two novel approaches are proposed to emulate the LISS III image with 740 km swath at 23.5 m spatial and 5-days temporal resolutions. First approach is to emulate the synthetic LISS III images at 23.5 m spatial and 5-days temporal resolutions. Mosaic such images to cover the full 740 km swath of AWiFS for a given date. First approach is achieved through a spatio-temporal data fusion technique which depends on the previously acquired single AWiFS-LISS III image pair. Second approach assumes that the non-overlapping region of AWiFS contains similar Earth’s surface features of LISS III overlapping region; then it is possible to enhance the spatial resolution of AWiFS to the spatial resolution of LISS III in the nonoverlapping region. It is achieved through a single-image-super resolution technique over Non-sub sampled Contourlet Transform. First approach is computationally efficient but it requires prior knowledge of a single AWiFS-LISS III image pair for each emulated LISS III image. That image pair is acquired before or after the prediction date. Also, first approach faces radiometric issues in the mosaic process. Second approach has high computational complexity. But it works well for the data sets which are satisfying the above basic assumption. An accuracy of both methods is validated with originally acquired LISS III data sets. Experimental results demonstrated that the accuracy of first approach is around 92 % and the second approach is around 87 %. In the second approach, only the overlapping regions of AWiFS and LISS III in simultaneous acquisition are used as prior knowledge. The accuracy of this method can be improved by increasing the database of the relevant prior knowledge.