Abstract: Geological information prediction and Earth monitoring by the satellite images are the recent research area to preserve the vegetation, weather forecast, and the disaster management. The employment of Hyper Spectral Image (HSI) by capturing the electromechanical energy variations from the Earth’s surface in the various spectral bands offers the significant contribution to the remote sensing applications. The clear image analysis depends on the spectral response. The capture of response in HSI in narrow bandwidth causes the less performance. Hence, the number of bands over the various time periods are the important requirement in clear image analysis. The multi-temporal images contain more information than RGB image since more bands are available in it. The absence of frames update leads to accuracy degradation. This study focuses on multi-temporal images for better isolation of normal and noise region and provides the clear image analysis compared to HSI. This study proposes the cellular automata-based noise filtering technique with the changes in noise prediction structure to eradicate the noise components, thereby better isolation is achieved. This study overcomes the update and accuracy limitations by an employment of image fusion to each band to eliminate the cloud and provide the necessary updated frames. The classification of normalized images from the fused images by using Tree Bagger algorithm with Neural Network (NN) formation (TBANN) predicts the cluster label for the color features of specific band results in the reduction of the atmospheric and signal dependent noise. The comparative analysis between the proposed TBANN with the existing methods regarding the accuracy, Kappa coefficient and the number of pixels count assures the effectiveness of TBANN in remote sensing applications.