Super-Resolution of LiDAR Data Using EDSR-CBAM Neural Networks

LiDAR data is crucial for various applications, including autonomous driving. Despite the growing interest in super-resolution techniques to improve LiDAR data quality, enhancing point cloud resolution remains challenging. This paper presents a new approach to LiDAR super-resolution, employing a hybrid neural network that combines the Enhanced Deep Super-Resolution (EDSR) and Convolutional Block Attention Modules (CBAM). Through spherical projections, our method achieves notable improvements in image quality metrics like PSNR and SSIM, as well as computational efficiency. Most importantly, our EDSR-CBAM model consistently outperforms existing U-Net models.