User:Asloob

Problems in Super-resolution
Super-resolution simply means to enhance the resolution of an imaging device. If someone cannot read a road sign board from some distance then one should move towards the sign board. This is a simple way one can resolve the unresolved data. In practical situation it is not possible to move the imaging device towards a scene to capture its high resolution images. A simple convex lens is an example of a simple imaging system. For a lens to act as a complete imaging system, something more is needed. For example a photo sensitive device which may be a simple photographic film or a charged coupled device (CCD) chip. The two basic elements of any imaging device are then a convex lens and a photo sensitive device. The resolution of an imaging system is therefore possibly limited either by a lens, or by a sampling device i.e. CCD or by both.

If the resolution of an imaging device is limited by an optical lens, then we talk about the classical limit or Rayleigh limit of resolution and the domain in which this problem of super-resolution falls is called Optical Super-resolution. Research work on optical super-resolution has been in progress since nearly four decades. If a convex lens is used to image a bright tiny spot of light, then we notice that in the image plane of the lens, the size of the spot has got bigger size than the actual spot in the object space. This may be because of diffraction, atmospheric turbulence or because of the abnormalities in the lens called aberrations and the mathematical term that takes care of this broadening in the image space is called the Point Spread Function (PSF) of the lens. A lens that acts close to an ideal lens has a very narrow PSF and a course lens has a broader PSF. Techniques to reduce the size of PSFs for improvement of resolution are one of the problems of super-resolution. There are a large number of techniques to estimate the PSF of an imaging lens and one of the techniques is called Blind De-convolution Technique.

Before the advent of CCDs, photo-sensitive materials used in the imaging processes were photographic films. An ordinary commercially available photographic film can safely resolve from 50 line pairs per mm to about 150 line pairs per mm where as a lens can resolve more lines per mm so the limit to resolution in an imaging system was due to poor resolution of the photographic films. Later big companies developed high resolution films that could resolve 3000 line pairs per mm to about 6000 line pairs per mm called Holographic films. The resolution capability of any good lens is always less than 3000 line pairs per mm. Scientists then started thinking to improve the resolution of an imaging lens rather than the films and a lot of techniques, hundreds in numbers, were developed but only a few of them are of practical nature.

In the conventional optical imaging systems the in-situ development of films has been a great problem and the advent of CCDs solved this problem. But CCDs posed another problem to Super-resolution and that was: The resolution is now limited by the CCD pixel size and not by the lens. This opened up another area of research to improve the limits of resolution imposed by the finite size of pixels in a CCD device. Recently some techniques have come up but there are always drawbacks associated with them.

Image processing techniques propose some solutions to super-resolution as well. Though such techniques are not reliable but some of them have produced really surprising results. The objective of this project is to invite students to work out the problems in super-resolution mentioned in the literature and to simulate some features of super-resolution problem. At least five to ten students can be accommodated on this project. This is a huge research area which will attract more students of different disciplines (Physics, Engineering, Computer Science) in the near future. If you really want to learn state of the art research in super-resolution then be a part of this project.