[ VIGRA Homepage | Class Index | Function Index | File Index | Main Page ]
![]() |
smooth.cxx | ![]() |
---|
Smooth an image using Recursive convolution functions functions: smooth.cxx
Usage: smooth infile outfile
/************************************************************************/ /* */ /* Copyright 1998-2002 by Ullrich Koethe */ /* Cognitive Systems Group, University of Hamburg, Germany */ /* */ /* This file is part of the VIGRA computer vision library. */ /* ( Version 1.3.3, Aug 18 2005 ) */ /* You may use, modify, and distribute this software according */ /* to the terms stated in the LICENSE file included in */ /* the VIGRA distribution. */ /* */ /* The VIGRA Website is */ /* http://kogs-www.informatik.uni-hamburg.de/~koethe/vigra/ */ /* Please direct questions, bug reports, and contributions to */ /* koethe@informatik.uni-hamburg.de */ /* */ /* THIS SOFTWARE IS PROVIDED AS IS AND WITHOUT ANY EXPRESS OR */ /* IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED */ /* WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. */ /* */ /************************************************************************/ #include <iostream> #include "vigra/stdimage.hxx" #include "vigra/convolution.hxx" #include "vigra/nonlineardiffusion.hxx" #include "vigra/impex.hxx" using namespace vigra; int main(int argc, char ** argv) { if(argc != 3) { std::cout << "Usage: " << argv[0] << " infile outfile" << std::endl; std::cout << "(supported formats: " << vigra::impexListFormats() << ")" << std::endl; return 1; } // Type of smoothing: int type; std::cout << "Type of smoothing (1 = Gauss, 2 = Exponential, 3 = nolinear) ? "; std::cin >> type; // input width of smoothing filter double scale; std::cout << "Amount of smoothing (operator scale) ? "; std::cin >> scale; double edge_threshold; if(type == 3) { std::cout << "Edge threshold ? "; std::cin >> edge_threshold; } try { vigra::ImageImportInfo info(argv[1]); if(info.isGrayscale()) { vigra::BImage in(info.width(), info.height()); vigra::BImage out(info.width(), info.height()); importImage(info, destImage(in)); switch(type) { case 2: { // apply recursive filter (exponential filter) to gray image recursiveSmoothX(srcImageRange(in), destImage(out), scale); recursiveSmoothY(srcImageRange(out), destImage(out), scale); break; } case 3: { // apply nonlinear diffusion to gray image nonlinearDiffusion(srcImageRange(in), destImage(out), vigra::DiffusivityFunctor<float>(edge_threshold), scale); break; } default: { vigra::FImage tmp(info.width(), info.height()); // apply Gaussian filter to gray image vigra::Kernel1D<double> gauss; gauss.initGaussian(scale); separableConvolveX(srcImageRange(in), destImage(tmp), kernel1d(gauss)); separableConvolveY(srcImageRange(tmp), destImage(out), kernel1d(gauss)); } } exportImage(srcImageRange(out), vigra::ImageExportInfo(argv[2])); } else { vigra::BRGBImage in(info.width(), info.height()); vigra::BRGBImage out(info.width(), info.height()); importImage(info, destImage(in)); switch(type) { case 2: { // apply recursive filter (exponential filter) to color image recursiveSmoothX(srcImageRange(in), destImage(out), scale); recursiveSmoothY(srcImageRange(out), destImage(out), scale); break; } case 3: { // apply nonlinear diffusion to color image nonlinearDiffusion(srcImageRange(in), destImage(out), vigra::DiffusivityFunctor<float>(edge_threshold), scale); break; } default: { vigra::FRGBImage tmp(info.width(), info.height()); // apply Gaussian filter to color image vigra::Kernel1D<double> gauss; gauss.initGaussian(scale); separableConvolveX(srcImageRange(in), destImage(tmp), kernel1d(gauss)); separableConvolveY(srcImageRange(tmp), destImage(out), kernel1d(gauss)); } } exportImage(srcImageRange(out), vigra::ImageExportInfo(argv[2])); } } catch (vigra::StdException & e) { std::cout << e.what() << std::endl; return 1; } return 0; }
© Ullrich Köthe (koethe@informatik.uni-hamburg.de) |
html generated using doxygen and Python
|