Datacube Preprocessing: Multitemporal Cross-Sensor Image Normalization

Multitemporal cross-sensor imagery provides fundamental data for Earth's surface monitoring over time. However, analyzing these images is a challenging task. Visual inconsistencies are often found in these images due to atmospheric and surface conditions variations. Hence image normalization in image pre-processing is considered to minimize the errors due to inconsistencies. There are several normalization methods that have been developed, such as histogram matching and linear regression by iteratively reweighted multivariate alteration detection (IR-MAD). Despite that, these methods have limitations in preserving important features and require a reference image that may not be available or adequately represent the target images. To address those limitations this study was conducted, a relaxation-based algorithm is proposed for Multitemporal cross-sensor satellite image normalization. This algorithm iteratively adjusts the radiometric values of images using a relaxation process. The effectiveness of this method was evaluated on multitemporal cross-sensor image datasets and showed significant improvements in radiometric consistency compared to other methods. The proposed relaxation algorithm outperformed IR-MAD and the original images in reducing radiometric inconsistencies, maintaining essential features, and improving the accuracy (MAE = 3.8; RMSE = 4.6) and consistency of surface reflectance values (R² = 94.01%).