We photographers often avoid small apertures to avoid loss of resolution from diffraction, and the diffraction problem often comes up when new full frame sensors with a high pixel density is discussed (even though the diffraction limit is not related to the pixel pitch). Previous threads in this forum have discussed how
superlenses can circumvent the diffraction limit through the use of lenses whose elements have a negative index of refraction. Unfortunately, photographic lenses with elements having a negative index of refraction are not yet available. Another approach to avoid the diffraction limit involves the use of
wavefront coding and image restoration using deconvolution. This
technology is already employed for cell phone cameras.
A recent
scientific paper describes how wavefront coding and post processing with a deconvolution filter can be used to break the diffraction limit in an optical system. Unfortunately, specially designed lenses (sometimes with a phase plate placed in the optical system) are required and the method is not ready for use with your Nikon or Canon camera lenses. Another problem with single lens reflex cameras is that the wavefront coded image appears blurred before it is post processed.
It is interesting to note that the effects of defocus and diffraction are related and often considered together, as discussed by David Jacobson in his
lens tutorial (see section V). It is reasonable to assume that methods to reduce defocus blur would also be useful in reducing diffraction blur, and it is possible that a diffraction limited image taken by an existing camera could be improved with a deconvolution filter such as FocusMagic or the Lucy-Richardson deconvolution. Determining the point spread function (PSP) necessary to undo the image degradation can be difficult, but sometimes an empirically determined PSP works reasonably well. Some photographers are already using FocusMagic or similar algorithms for capture sharpening. Has anyone tried such methods on diffracted images taken with small apertures? In this case, I would expect that a high pixel density such as in the 1DsMIII would provide a better image for deconvolution.
Bill
