L0-gradient-smoothing

---

This is a C++ implementation of "Image Smoothing via L0 Gradient Minimization", Li Xu, Cewu Lu, Yi Xu, Jiaya Jia, SIGGRAPH ASIA 2011.

Dependencies:

The code depends on following libraries.

• OpenCV 2.4.9
• Eigen 3.2
• Boost 1.55.0

Build:

You can build the code using CMake.

mkdir build	
cd build
cmake ..
make 

Usage:

usage: L0-gradient-minimization [-h] [-i input_img] [-o out_dir] [-c cofig_name]

arguments:
 	-h, --help       show help message
 	-i, --input	     input image filename 
 	-o, --output     output path
 	-c, --config     config filename

Parameters in the algorithm can be controlled through a config file. Config file must include following entries. See config_sample.txt.

lambda:     smoothing parameter
beta_max:   parameter of the termination of the algorithm
kappa:      parameter relating to the speed of the algorithm
exact:      solver type for linear system. if true LU factorization is used , else Conjugate Gradient method is used

Example:

input

output(λ=0.005)

output(λ=0.03)