Low-Resolution Surveillance Face Image Denoising Using Multistage Multifilter Algorithm

Face recognition in surveillance video depends on the surveillance camera and the surrounding environment. The problem becomes more complex to identify human faces in such surveillance videos when the frames have low resolution details and are affected by various noises. While several researchers have attempted to remove noise from degraded video frame images, there remains a void in the research of denoising surveillance images. It is observed that the existing denoising algorithms are not giving satisfactory results. Hence, we propose a Multistage Multifilter (MSMF) algorithm with four stages which employs super pixel segmentation followed by three different denoising techniques, namely Principal Component Analysis (PCA), Median, and anisotropic diffusion filtering, to denoise the low-resolution and noisy surveillance images. Overall, this approach to denoise surveillance images involves a combination of statistical, nonlinear, and edge-preserving filtering techniques to address the challenges posed by the noise present in these low-resolution images. The results are quite promising with 59.12% as the average Peak Signal to Noise Ratio (PSNR) value for Gaussian Noise and 60.15% as the average PSNR value for Salt and Pepper Noise.