3D Gaussian SLAM with DoG Mapping

The goal of this study is to address blur artifacts and inconsistently tracked depth data in conventional RGBD Simultaneous Localization and Mapping(SLAM) systems. To tackle these issues, we propose a method that uses the Difference of Gaussian(DoG) to detect high-frequency regions, enhancing scene details. Experimental results show that our method achieves higher Peak Signal-to-Noise Ratio(PSNR) and lower Average Trajectory Error(ATE) compared to existing approaches. Even though the modified loss function was applied only during the mapping phase, localization performance also improved. These results indicate that our method can enhance the accuracy and reliability of dense SLAM systems.