Gaussian Fourier Pyramid for Local Laplacian Filter

The optimized code for the paper is published in the part of my library OpenCP. The detail part for this paper is here.

However, the code published on OpenCP is heavily optimized, i.e., less readable. Therefore this page also provides readable codes: C++ header only code and Python implementation based on OpenCV. The header only code has about 500 lines, while the optimized code has 15000+ lines.

If you need faster operation, please use the code in the OpenCP project.

• header only C++
  • The sln file (GaussianFourierPyramid.sln) is for the project of header only C++.
• Python
  • under construction.

Usage (C++)

1. Generate an instance (FastLLF llf or GaussianFourierLLF gfllf)
2. Generate parameter maps if you need adaptive filtering
3. Call the filter method (llf.filter, gfllf.filter)

The sample code is as follows:

    //load image
	Mat src = imread("flower.png");
	//destination image
	Mat destFastLLF, destFourierLLF, destFastLLFAaptive, destFourierLLFAaptive;
	
	//parameter setting
	const float sigma = 30.f;
	const float boost = 2.f;
	const int level = 2;
	const int order = 4;

	//create instance
	FastLLF llf;
	GaussianFourierLLF gfllf;

	//parameter fix filter
	llf.filter(src, destFastLLF, order * 2, sigma, boost, level);//order*2: FourierLLF requires double pyramids due to cos and sin pyramids; thus we double the order to adjust the number of pyramids.
	gfllf.filter(src, destFourierLLF, order, sigma, boost, level);
	
	//parameter adaptive filter
	//generate parameter maps (circle shape)
	Mat sigmaMap(src.size(), CV_32F);
	sigmaMap.setTo(sigma);
	circle(sigmaMap, Point(src.size()) / 2, src.cols / 4, Scalar::all(sigma * 2.f), cv::FILLED);

	Mat boostMap(src.size(), CV_32F);
	boostMap.setTo(boost);
	circle(boostMap, Point(src.size()) / 2, src.cols / 4, Scalar::all(boost * 2.0), cv::FILLED);

	//filter
	llf.filter(src, destFastLLFAaptive, order * 2, sigmaMap, boostMap, level);
	gfllf.filter(src, destFourierLLFAaptive, order, sigmaMap, boostMap, level);

Information

Y. Sumiya, T. Otsuka, Y. Maeda, and N. Fukushima, “Gaussian Fourier Pyramid for Local Laplacian Filter,” IEEE Signal Processing Letters (SPL), vol. 29, pp. 11-15, 2022.

• IEEE Xplore

@article{sumiya2022spl,
    author  = {Y. Sumiya and T. Otsuka and Y. Maeda and N. Fukushima},
    title   = {Gaussian Fourier Pyramid for Local Laplacian Filter},
    journal = {IEEE Signal Processing Letters},
    volume  = {29},
    number  = {},
    pages   = {11-15},
    year    = {2022},
}