Call recovery.solve_grid_linear_simple3() or recovery.solve_grid_linear_simple3_solver(). You can look at recovery.test_solve_grid_linear_simpleN() for an example. If you run python recovery.py it will call solve_grid_linear_simple3() and generate test output.

Look into modifying surface_with_edges.py main(). It calls recovery.solve_grid_linear(). (It relies on lots of external helper function to load its input, which isn't necessary.)

To smooth bumps created by normals constraints, try calling recovery.smooth_bumps() or recovery.smooth_bumps2() with the locations of the bumps and some integer value for bump_radius.

To replace pixels in an image whose opacity is not 100% with a harmonic function:

    python fill_alpha_harmonic.py path/to/image_with_non_opaque_pixels.png path/to/output.png

100% opaque pixels remain unchanged.

• Python >= 2.6 and < 3
• PIL (Python Image Library) (only needed for the examples and tests)
• scipy
• cvxopt