Vignetting Correction

The code is implemented according to paper of Single-Image Vignetting Correction*,by Yuanjie Zheng,Stephen Lin,etc.Both c++ and Matlab source codes are provided. For the Matlab version, I revise some parts of the code offered by the author, Yuanjie Zheng. For the C++ version, it is implemented by my supervisor, Donghai Xie and me.

C++ version:

Before compiling the source code, please make sure you have configured the OpenCV open source library properly. For C++ version, there are 2 projects. Vignetting project is used to produce vignet for images. Vignetting correction project is used to correct the vignetting on the images.

Matlab version:

Copyright (c) 2010, Yuanjie Zheng All rights reserved.
For the Matlab version, you can run the demo_vignetting module to start the correction process. There is also another module called Estimate, which is used to estimate the performance of the correction. See Correction Estimate part or the paper material for details.

Results

Lena

Corrected

Sea

Corrected

Couple

Corrected

Performance Estimate

We estimate the performance from 2 perspectives. First, since we can produce the vignetting function through the C++ project Vignetting, we can compare the vignetting function defined by ourselves with the vignetting function we calculate through the correction process.

Real Vignetting Function

Estimate Function

If we fit value along the diagonal of the 2 images above, we can get a result like this.

Fit

Second, according to the paper, the radius gradient distribution of image without vignetting should be centrosymmetry. Thus, we can calculate the value of log(1+|x|),where x represents the radius gradient value, to see whether the 2 sides of the distribution overlap with each other. And we can get the result like this.

Before Correction

Corrected Result