Light Field Reconstruction Using Shearlet Transform.

Area of Science:

• Computer Vision
• Image Processing
• Computer Graphics

Background:

• Image-based rendering (IBR) techniques aim to synthesize novel views of a scene from a limited set of input images.
• Reconstructing dense light fields is crucial for various applications, including virtual reality and 3D content creation.
• Existing methods often struggle with large view disparities and semi-transparent objects.

Purpose of the Study:

• To develop an advanced image-based rendering technique for high-quality light field reconstruction.
• To leverage sparse representation in the shearlet transform domain for improved reconstruction accuracy.
• To enhance the rendering of challenging scenes, particularly those with semi-transparent elements.

Main Methods:

• Utilizing a modified shearlet transform specifically adapted for epipolar-plane images (EPI).
• Implementing an iterative regularization algorithm with adaptive thresholding for robust reconstruction.
• Processing sparse representations of EPIs to reconstruct densely sampled light fields.

Main Results:

• Achieved high-quality light field reconstruction even with significant disparities between camera views.
• Demonstrated superior performance compared to state-of-the-art depth image-based rendering (DIBR) techniques.
• Successfully reconstructed scenes containing semi-transparent objects, a common challenge for existing methods.

Conclusions:

• The proposed shearlet transform-based light field reconstruction method offers significant improvements in rendering quality.
• The technique is suitable for a wide range of applications requiring dense light field data.
• This approach provides a robust solution for rendering complex scenes, advancing the field of image-based rendering.