Perbandingan Kinerja Filter Citr Evaluasi Filter Spasial untuk Reduksi Noise dan Deteksi Tepi pada Citra CCTV

Abstract

Digital images generated by Closed Circuit Television (CCTV) cameras play an important role in security systems. However, these images often experience quality degradation due to noise interference, complicating the identification process. Therefore, this study aims to compare the effectiveness of three spatial filtering methods: Gaussian Filter, Median Filter, and Sobel Filter. The comparative experiment uses the Python programming language and OpenCV library. The testing applies low-pass filters (Gaussian and Median) to reduce noise, and a high-pass filter (Sobel) for edge detection. Image quality is evaluated objectively using Mean Squared Error (MSE) and Peak Signal to Noise Ratio (PSNR) parameters. The results indicate that filtering performance depends heavily on the type of noise. The Median Filter is superior in reducing impulsive noise while preserving details, whereas the Gaussian Filter is optimal for normally distributed noise. Meanwhile, the Sobel Filter is highly effective in extracting object boundary lines after the image smoothing stage. In conclusion, selecting the appropriate filtering method based on noise characteristics is crucial to produce optimal surveillance image quality.