Depth Map / Defoucs Map from Single Image - GPU

Is anyone aware of simple algorithm to extract Depth Map / Defocus Map from a single regular image.

I found some algorithms, yet most of them require "Opt开发者_StackOverflow中文版imization" step which isn't suitable to GPU's.

It doesn't have to be 100% perfect, just good enough.

P.S. Defocus Map - Greyscale Map of the "Blur" level of each pixel.

Calculating depth from a single image is inherently ill-conditioned. There will always be user input or assumptions about the scene required. These assumptions/inputs are used in optimization algorithms. I think this is your best option, because these will give the best results.

If you want to use optimization on (modern) GPU's, you can use something like SpaXe or CUDACuts. It's not impossible, but the usefulness depends on the application.