A Modified Singular Value Decomposition (MSVD) Approach for the Enhancement of CCTV Low Quality Images

This Project proposes an efficient method for enhancing low-quality CCTV video frames using a Modified Singular Value Decomposition (MSVD) technique. The process begins with the acquisition of input videos from a university database, which are recorded at a resolution of megapixels. These videos are then converted into 2D frames with a standardized resolution of pixels using MATLAB, enabling consistent image processing. Singular Value Decomposition (SVD) is applied to decompose each frame into three matrices: left orthogonal, sigma, and right orthogonal matrices. For enhancement, the diagonal sigma values of low-quality frames are replaced with those from good-quality frames, effectively transferring quality features to degraded images. The final step involves reconstructing enhanced images from the modified SVD matrices and evaluating their performance using six key metrics: Peak Signal-to-Noise Ratio (PSNR), Entropy, Mean Squared Error (MSE), Contrast, Time Computation, Structural Similarity Index (SSIM). This approach not only enhances visual clarity but also ensures improved performance metrics, demonstrating the effectiveness of MSVD in CCTV video enhancement tasks. The proposed method is particularly useful for applications requiring high-quality visual outputs from surveillance systems, making it a valuable tool for modern video processing applications.