Pipeline de processamento de imagens com baixa luminosidade e métodos cegos de avaliação

Abstract

This workpresentsthedevelopmentofanimageprocessingpipelineaimedatenhancing photographs captured in low-light environments. The proposal is based on the use of RAW images, which preserve the complete information from the camera sensor, allowing for greater flexibility in post-processing. The designed pipeline is composed of a set of classical Image Signal Processing (ISP) stages, including RAW image reading and linearization, demosaicing, color correction, white balance, exposure adjustment, gammacorrection, haze removal, local contrast enhancement (CLAHE), noise reduction, and saturation adjustment. The quality of the processed images was evaluated using no-reference objective metrics, such as BRISQUE and NIQE, as well as qualitative visual inspection. The results show that the proposed pipeline is effective in recovering dark regions, improving sharpness, and revealing previously hidden details. Comparisons between scenarios with and without noise filtering provided additional insight into the impact of this step on perceived image quality. The system is modular, interpretable, and suitable for various low-light capture contexts.