Data graph classification and super-resolution reconstruction methods based on deep learning

During the process of structured light 3D reconstruction, the resolution of images plays a crucial role in the final reconstruction outcome. In practical terms, there are primarily two methods to enhance image resolution. Firstly, improving the performance of the camera can directly provide higher-resolution images, albeit at a higher cost. Secondly, employing deep learning methods for preprocessing acquired images before using 3D reconstruction algorithms offers another viable approach. In handling image processing practically, to address issues related to hardware performance due to excessively high actual resolutions, a novel classification method is employed for distinguishing and categorizing effective regions within the images. Following the classification and identification of these effective regions, super-resolution processing is applied specifically to these areas. This approach reduces the time required for super-resolution processing and mitigates the hardware demands of deep learning.