SIFT density image function and filter
The idea here is to code an image function (similar to the variance image function, for instance) which computes the number of SIFT/per pixel detected in the neighborhood of a given pixel in a scalar image. This function can afterwards be used in a filter in order to produce an image of densities.
The SIFT image function can be seen as a particular case where a detector (producing points, lines etc., that is vector data) + a count method (simple count, count + condition on the detected objects, density -- count/nb pixels) are combined to produce meaningful information:
CountImageFunction< DetectorType, CountStrategy, ConditionOnObject=AlwaysTrue>
ConditionOnObject is for instance a functor which returns a boolean value after operating on single detections produced by the DetectorType. CountStrategy can be SimpleCount, PerPixelDensity, etc.