Automatic 3D View Generation Using Super-Resolution and Automatic Alignment from Uncalibrated Stereo Image Pairs
Jamal Sami Zraqou, Adnan H. Al-Helali, Wissam Alkhadour, Jawad Hasan Alkhateeb · 2025
One of the most important methods to create a stereoscopic 3D image is called Anaglyph generation; it uses two images to get the scene's depth. This work proposes a new method to perform face hallucination and use auto-detection for point matching to cater to the performance differences in input images. Drawing upon three principal techniques that effectively work around two principal difficulties of prior approaches, resolution and misalignment, this contribution provides a smoother Anaglyph image. The quality of the generated Anaglyphs is evaluated qualitatively. The first Anaglyphs are considerably less clear and have worse depth estimation than the final ones. Besides, the automation process has no restriction on the orientation of the cameras and is better suited for Anaglyph creation than ordinary approaches.