Image Restoration-Theory and Methods


What is image restoration in digital image processing? 

Digital images are used to display or record useful information. But there are occasions when the produced images represent a degraded image apart from the original version. It is mainly due to imperfections in the image capturing process.

The undoing of these flaws are critical to most of the succeeding image processing works. There exists a broad variety of diverse degradations including instance noise, illumination, geometrical degradations and color imperfections.  In digital image processing, image restoration refers to a line of methods to reduce this degradation. The image degradation may happen:

  • During the display mode
  • Acquisition mode
  • Due to sensor noise
  • During processing mode
  • Blur that result from camera focus issue
  • Random turbulence of the atmosphere
  • Object-camera motion problems

Image Restoration

Most of the existing image restoration methods we begin with an assumption that the degradation procedure can be defined using a mathematical model. The success of this model depends on how deeply we are aware of the original image characteristics and the level of degradation that happened to the image under consideration.

What is the difference between image enhancement and restoration? Image enhancement is related to extraction or accentuation of the image features whereas image restoration deals with degradation elements associated with an image.

An imaging system consists of an image formation system, recorder and a detector. We can outline such a system using a general model which can be defined as:


Image degradation process- general model:

A basic version for the picture restoration process model is:Y(i,j)=H[f(i,j)]+n(i,j)

Where y(i, j) represents the degraded image, f (i, j) the original image, n(i,j) is the image independent external noise and H is an operator that represents the image degradation process. The image restoration methods can be classified as:

  • Stochastic
  • Deterministic
  • Blind
  • Non-blind
  • Semi-blind