Acquisition of satellite imagery, frequently becomes corrupted by various forms of noise, which degrades image quality and impedes accurate interpretation and subsequent analysis. This paper presents a novel, integrated approach for effective noise removal from such imagery, utilizing the capabilities of the Wavelet transform based thresholding, Singular Value Decomposition (SVD), and also autoencoders. The Wiener filter, proves adept at addressing stationary noise, particularly effective in suppressing additive Gaussian noise whilst preserving important details based on local statistics. Following this, SVDs are employed to exploit the inherent low-rank structure characteristic of natural images, enabling further noise reduction by reconstructing the image from its more significant singular values, thereby separating signal from noise components. Finally, an autoencoder, which is are powerful deep learning architecture, is then utilized for its capacity to learn complex, non-linear representations from noisy inputs. This data-driven strategy allows the autoencoder to effectively apprehend intricate noise patterns and reconstruct high-fidelity images, offering particular benefit for complex or non-stationary noise types where traditional filters may struggle. By combining these complementary techniques, the proposed methodology aims for superior noise suppression, the preservation of crucial image features, and the enhancement of overall visual quality and analytical utility of satellite images. Expected results demonstrate significant improvements in signal-to-noise ratio and perceptual quality compared to individual methods, highlighting the synergistic benefits of this combined framework.