Depth from focal gradient analysis using object texture removal by albedo normalization

DETAILED DESCRIPTION The method and apparatus of the invention substantially remove the effects of object reflectance from an image acquired using structured illumination, so that only the structured illumination pattern and its degree of focus/defocus remains.
Removing the effects of object reflectance (the removing called "albedo normalization") is achieved by dividing an image of an object taken under structured illumination by a corresponding image of the object taken under uniform illumination.
The albedo normalization act removes the primary source of noise in range images obtained using a depth from defocus or depth from focus of structured illumination technique, by removing spurious image frequencies from the image before processing by a focus measure.
The albedo normalization act permits the depth from defocus and depth from focus techniques to be used for one or more focal positions, over a broad range of materials of interest in machine vision.
In one general aspect, the invention is a method for providing a range image of an object, where the method includes the acts of acquiring a structured illumination image of the object; acquiring a uniform illumination image of the object; performing albedo normalization using the structured illumination image of the object and the uniform illumination image of the object to provide an albedo-normalized image; processing the albedo-normalized image using a focus measure to provide a focal image; and using the focal image to provide a range image.
In a preferred embodiments, the act of processing the albedo-normalized image using a focus measure to provide a focal image includes the act of using an RMS operation, and the act of processing the albedo-normalized image using a focus measure to provide a focal image includes the act of using a Laplacian-like convolution.
In further preferred embodiments, the act of using the focal image to provide a range image includes the act of using a table look-up calibration employing a look-up table, and the act of using a table look-up calibration includes the acts of collecting focus measure data at known Z-positions, and inverting a look-up table using interpolation