Generation of NDVI Time Series using a Hybrid Regression Kalman Filter based Approach

Abstract

The synthetic satellite images can be generated by various methodologies using available Landsat images and the MODIS composite. This paper uses a hybrid methodology combining regression analysis, Kalman filtering, and smoothing. It combines the forward recursion Kalman filter with the backward recursion Kalman filter, which is named a combined mode Kalman filter. This improved hybrid technique provides more accurate synthetic satellite images than those provided by the other blending algorithms like STARFM, ESTARFM, SPSTFM, and KFRFM. Residuals are lower for the combined recursion for the generated synthetic NDVI image generated by the forward or backward recursion filter.