Public CCTV image quality in Yogyakarta's tourism areas often suffers from low resolution, complicating automatic vehicle license plate identification. This study contributes an empirical comparative evaluation of classic interpolation methods (Bicubic Interpolation and Lanczos4 In-terpolation) applied to real-world public CCTV traffic scenarios. High-Resolution (HR) reference images were downsampled by 4x to simulate low-resolution (LR) CCTV footage, then upscaled back using both methods. Performance was quantitatively measured using Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index (SSIM). Experimental results on 10 sample images show that Bicubic Interpolation achieves a slightly higher average performance (PSNR 19.47 dB and SSIM 0.5325) compared to Lanczos4 (PSNR 19.25 dB and SSIM 0.5128). However, qualitative visual analysis confirms that Lanczos4 produces sharper character edges. Nevertheless, as this qualitative enhancement has not been quantitatively validated through Optical Character Recognition (OCR) testing or statistical analysis, these conventional interpolation methods serve pri-marily as preliminary preprocessing steps for potential integration into automatic recognition systems.
Copyrights © 2026