New pixel-based methods for low-light image enhancement

Low-light image enhancement is one of the most important research topics in the field of image processing. Images captured in low-light environments exhibit artifacts that affect both human perception and computer vision applications due to issues such as low contrast, color distortion, and severe noise. Therefore, low-light image enhancement plays a critical role in advanced image processing tasks. The aim of low-light image enhancement is to improve contrast, illumination, and detail information in low-light images without introducing color distortion or noise amplification. In this thesis, four different low-light image enhancement methods are proposed: "Sharpening-Smoothing Image Filter based method", "Metropolis Theorem based method", "New Color Channel Driven Physical Lighting Model based method", and "Adaptive Chrominance Gamma Correction based method". The proposed methods are explained in detail with regard to their theoretical background and the strategies they employ. Comparative analyses, both visual and quantitative, are conducted on frequently used datasets in the literature to clearly demonstrate their strengths and limitations. The findings reveal that there is no universal optimal solution, and that each proposed method becomes more effective in different low-light scenarios. The evaluation results indicate that the low-light image enhancement problem is highly scene-dependent. Therefore, it is evident that different approaches should be adopted for different applications. This thesis comprehensively analyzes the proposed methods for low-light image enhancement, highlighting their strengths and limitations, and also provides insights into potential future work in this field.