An Efficient Image Processing Approach for Cloud Removal in Satellite Images

Abstract

Cloud cover can substantially restrict the information available in optical satellite images, affecting applications including agricultural monitoring, disaster assessment, and urban planning. Although deep-learning and multi-temporal techniques have improved cloud mitigation, they commonly depend on considerable computational resources, training data, or repeated observations. This study presents a classical image-processing framework for identifying cloud-covered regions and restoring the affected image areas. The proposed method uses CLAHE (Contrast Limited Adaptive Histogram Equalization)-based contrast enhancement, HSV (Hue, Saturation, Value)-based masking, connected component analysis, and a two-step inpainting algorithm. The proposed method is evaluated on the RICE dataset using mean squared error, root mean square error, peak signal-to-noise ratio, structural similarity index measure, accuracy, and area under the receiver operating characteristic curve. The resulting framework provided effective cloud detection while retaining a relatively lightweight computational requirement. Although the reconstruction quality remains lower than that of recent deep-learning approaches, the method provides an efficient baseline for resource-constrained environments where computational efficiency is prioritized.