A denoising method for inter-annual NDVI time series derived from Landsat images

摘要

The inter-annual Normalized Difference Vegetation Index (NDVI) time series derived from Landsat (INSL) has been applied to many fields. However, INSL may contain noise from many factors. At present, most researchers focus on denoising for NDVI time series derived from other datasets with shorter time intervals. In this paper, we proposed a denoising method for INSL by combining the best index slope extraction algorithm (BISE) with wavelet transform (WT). This method consists of two steps: (a) INSL are processed by BISE under a limiting condition to remove the significant noise; and (b) INSL obtained by step (a) are filtered by WT. Commonly used methods, including Savitzky-Golay filter (S-G) and BISE, were used as references to evaluate the proposed method. Using qualitative and quantitative comparisons and a change detection analysis, the proposed method was demonstrated to be reliable.