Reconstruction of 3D Surfaces with Complex Material Composure Using a Light Field Camera
- Publication Type:
- Thesis
- Issue Date:
- 2021
Open Access
Copyright Clearance Process
- Recently Added
- In Progress
- Open Access
This item is open access.
Representing real-world objects on a digital screen is a significant and challenging topic in the area of computer vision and augmented reality. This work addressed the challenge of reconstruction of 3D surfaces with a complicated material appearance by using a light field camera. Most recent research uses single images to address this problem, but without using a light field camera, encounter difficulties and limitations to overcome this problem. However, we show that by using a light field camera without user interaction or any requirement for object planarity or symmetry, reconstruction of a 3D model with high accuracy is possible. A light field camera, also known as a Plenoptic camera can capture rich information about the spatial and angular distribution, as well as intensity and colour of light in a single shot. For the reconstruction of 3D models, creating a 3D point cloud is essential and is often obtained based on a depth map. As a result, first, we developed a robust method to estimate an accurate depth map based on the combination of sub-aperture image matching and defocusing cues for a 4D light field format. The depth map is refined using a fast-weighted median filter providing robustness to noise. In the second part, we proposed a novel strategy for the creation of a 3D point cloud from the depth map of a single 4D light field image. The proposed method is based on the transformation of point-plane correspondences. Considering the estimated depth map from the previous part, we applied histogram equalization and histogram stretching to enhance the separation between depth planes. In the third step, we improved our suggested method to obtain a dense and more accurate three-dimensional (3D) point cloud. We applied intelligent edge detection by using feature matching and fuzzy logic from the central sub-aperture light field image and the depth map. The results showed that our new method can reliably mitigate noise compared to other existing methods. Finally, having obtained the 3D point cloud we handled the problem of reflectance in complex material appearance. We developed a new strategy to recover reflectance information based on colour analysis as well as brightness analysis of a light field image. Experimental results demonstrate the effectiveness of our method in both synthetic and real-world images compared to other states of the art methods. Overall, 3D reconstruction can solve many problems of computer vision that is still a challenging topic.
Please use this identifier to cite or link to this item: