vincentqyw/depth-estimation-light-field

👾 A collection of code of depth estimation using light field data.

Depth Estimation Light Field
A collection of light field depth estimation algorithms

This is a repository hosts some depth estimation algorithms using light field. Feel free to add more!

Usage

git clone --recurse-submodules https://github.com/Vincentqyw/Depth-Estimation-Light-Field.git

LF_DC

Source code used in the following paper. Michael W. Tao, Sunil Hadap, Jitendra Malik, and Ravi Ramamoorthi. Depth from Combining Defocus and Correspondence Using Light-Field Cameras. In Proceedings of International Conference on Computer Vision (ICCV), 2013.

LF

Source code used in Accurate Depth Map Estimation from a Lenslet Light Field Camera, CVPR 2015

CAE

CAE refers to Constrained Angular Entropy Cost. This is the source code used in

1. Robust Light Field Depth Estimation for Noisy Scene with Occlusion, CVPR 2016
2. Robust Light Field Depth Estimation using Occlusion-Noise Aware Data Costs, PAMI 2017

Homepage: Robust Light Field Depth Estimation using Occlusion-Noise Aware Data Costs

SPO

SPO refers to Spinning Parallelogram Operator. PAPER TO CITE: Shuo Zhang, Hao Sheng, Chao Li, Jun Zhang and Zhang Xiong. Robust depth estimation for light field via spinning parallelogram operator, Computer Vision and Image Understanding, 2016, 145(C), 148-159

depth-estimation-light-field's Issues

Depth for adjacent views in a light field

Hi, thank you for this repository! I've been looking for a light field dataset which provides depth map for each adjacent view, like in a usual stereo setting, but I've only found datasets/methods that provide a more consolidated depth map using multiple views, which can not be seen from the usual 2-view stereo perspective. Do you know of any method/dataset which might help in this direction?

I tried getting the depth map using various existing algos/models, but because the adjacent views usually have very low disparity the results are mostly unusable. Looking forward to your suggestion!

在自己的数据集上测试

请问为什么我在这些深度估计算法上测试使用自己的illum相机拍摄的图片时，都无法取得令人满意的效果，在输入真实世界图片时有什么注意事项吗？