Homography Propagation and Optimization for Wide-baseline Street Image Interpolation
Demo code for the work below:
- Nie, Y., Zhang, Z., Sun, H., Su, T., & Li, G. "Homography Propagation and Optimization for Wide-baseline Street Image Interpolation." IEEE Transactions on Visualization and Computer Graphics, 23(10), 2328-2341, 2017.
Video Demo on YouTube.
External libraries and code used:
- Boost 1.62: https://www.boost.org/
- OpenCV 2.49: https://opencv.org/
Run the code
- The code is currently tested on VS 2012.
- Configure the data path in the main.cpp file.
- Define __FORWARD in nrdc_processing.h to run forward warping. The warping result will be saved in data/viewinterp/1.
- Define __BACKWARD in nrdc_processing.h to run backward warping. The warping result will be saved in data/viewinterp/2. Note: DO not define __FORWARD and __BACKWARD at the same time.
- Run the Matlab code in the folder multibandblending to render the final result, which will be save in data/viewterp/3.
Common Questions
Q1: How to generate the labels.txt file in the forward/backward folder?
A1: This file provides an initial segmentation of the image. Any reasonable superpixel segmentation algorithm (For example, SLIC) can be used.
Q2: How to generate the matches.txt file in the forward/backward folder?
A2: This file can be generated from any ANNF algorithms, for example, PatchMatch, DAISY Filter Flow and NRDC.
Citation:
@article{nie2017homography,
title={Homography propagation and optimization for wide-baseline street image interpolation},
author={Nie, Yongwei and Zhang, Zhensong and Sun, Hanqiu and Su, Tan and Li, Guiqing},
journal={IEEE transactions on visualization and computer graphics},
volume={23},
number={10},
pages={2328--2341},
year={2017},
publisher={IEEE} }
Code by Zhensong Zhang and Yongwei Nie. For research purpose ONLY.