Image Enhancement on Surveillance Footage Using Enhanced Deep Residual Network Super Resolution

The study presents an approach for pedestrian detection from CCTV footage with the application of Enhanced Deep Residual Network Super Resolution (EDSR), in addition to YOLOv8 for person detection, to overcome blurry and low- resolution scenarios through image enhancement. For performance evaluation, Peak Signal-to-Noise Ratio (PSNR) and Mean Square Error (MSE) are used to look at the pedestrian detection and super resolution performance, and compared to several other traditional enhancement methods namely bicubic, bilinear and nearest neighbor interpolation. Our positive results affirm the method's potential to be used in future enhancement tasks.