Computational photography experiments

Algorithm outline

  • Convert the input RGB image to YCbCr color space.
  • Compute the gradient of the luminance channel (Y).
  • Calculate a new luminance channel by solving the partial differential equation of Poisson. This is done using a Gauss-Seidel solver at the moment, and this is why it doesn't converge at speeds that are sufficient for realtime use.
  • Convert back to RGB color space and write the result image into your favorite format.
  • Result

    The original photograph is on the left, and the result is shown on the right side. This is a photo I took from Rovaniemi, Finland.