Documentation is a bit hazy on the subject…
Is the energy map input supposed to be a map that gets evaluated by the tool to define where the lowest energy paths are, or is it supposed to contain a map of the energy gradients or path ranks themselves?
So far, every test I’ve run has failed, so I’m thinking I need to really fine tune the energy input, but I’m not sure what I can do to improve the results.
I agree the documentation is a bit short for the beta plugins. I’ll try and expand on it today.
Essentially the best base energy for the content aware resize is a Sobel or Laplacian image. If you use the Fusion native Sobel Filter then the output seems to be normalized, so you have to disable the “Make White energy pixel infinite” option. Once you have the base energy you can add a paint tool on top to mask out areas.
Brian!
But is the idea that you are feeding it the gradient magnitude? And you don’t break it out for X and Y, but use one image (Value? Luminance?) for both?
The x and y are combined for the gradient magnitude. If they are not combined then the lowest energy path sneaks through the openings. The energy of a pixel is defined as the maximum value from the R, G, or B of the color channel in the energy image.
Brian!