High-spatial resolution Remote Sensing (RS) data provides detailed information which ensures high-definition visual image analysis of earth surface features. These data sets also support improved information extraction capabilities at a fine scale. In order to improve the spatial resolution of coarser resolution RS data, the Super Resolution Reconstruction (SRR) technique has become widely acknowledged which focused on multi-angular image sequences. In this study multi-angle CHRIS/PROBA data of Kutch area is used for SR image reconstruction to enhance the spatial resolution from 18 m to 6m in the hope to obtain a better land cover classification. Various SR approaches like Projection onto Convex Sets (POCS), Robust, Iterative Back Projection (IBP), Non-Uniform Interpolation and Structure-Adaptive Normalized Convolution (SANC) chosen for this study. Subjective assessment through visual interpretation shows substantial improvement in land cover details. Quantitative measures including peak signal to noise ratio and structural similarity are used for the evaluation of the image quality. It was observed that SANC SR technique using Vandewalle algorithm for the low resolution image registration outperformed the other techniques. After that SVM based classifier is used for the classification of SRR and data resampled to 6m spatial resolution using bi-cubic interpolation. A comparative analysis is carried out between classified data of bicubic interpolated and SR derived images of CHRIS/PROBA and SR derived classified data have shown a significant improvement of 10-12% in the overall accuracy. The results demonstrated that SR methods is able to improve spatial detail of multi-angle images as well as the classification accuracy.

Date Published: 30 April 2016
PDF: 12 pages
Proc. SPIE 9880, Multispectral, Hyperspectral, and Ultraspectral Remote Sensing Technology, Techniques and Applications VI, 98800Y (30 April 2016); doi: 10.1117/12.2227665

Published in SPIE Proceedings Vol. 9880:
Multispectral, Hyperspectral, and Ultraspectral Remote Sensing Technology, Techniques and Applications VI
Allen M. Larar; Prakash Chauhan; Makoto Suzuki; Jianyu Wang, Editor(s)