Introduction to Digital Image Processing

מאמר שלי שפורסם בעיתון עילוי של מנסה ישראל. גיליון 9, תשרי התשס"ה / אוקטובר 2004

Since the great quest for knowledge began, many tools have been created to gather as much useful information as possible. Nowadays, in the field of image processing, special means have been designed to extract more information from the data gathered by the variety of visualization tools.

Welcome to the field of Digital Image Processing

Digital Image Processing generally refers to the processing of a two dimensional picture by digital means (such as a computer), where an obtained image is transformed from a continuous scene F (X,Y) into digital image f[x][y] that is the discrete representation o F (X,Y). Each cell in the two-dimensional structure is referred to as a Pixel; the pixel contains the different parameters which determine its appearance.

Digital Image processing has a broad spectrum of applications, starting from medical systems, such a low radiation x-ray pictures and endoscopy medical imaging, to remote sensing via satellites, for example meteorological or military applications.  The images acquired by these means are extremely useful for supplying various kinds of information.

As has been previously mentioned, the problem often is not getting the images, but extracting the needed information. For example, the satellite images, which are usually half-tone normal black/white pictures, often suffer from distortions such as Grain and Blur caused by various reasons like atmosphere, technical problems, lack of focus and even data limitations, However, not only should we deal with these problems, but also we should do it fast, even very fast! For the simplest satellite can send up to 100,000 images from a single location! So whatever the solution is, it must have low computational cost. After enhancing these images, they are analyzed by a computer and stored, which represents a storage problem and require high data comparison.

Medical image processing had quite similar problems, but must have different solutions. Limitations that were mentioned before such as low computational cost and high data compression are not an option when it comes to human life. Whereas a satellite image can be brutally compressed up to 36% from its original size and can be saved as a low-resolution image, a medical image must be handled with extreme care and will be saved in more than 1 GB. (Compression is not very effective as an image usually keeps up to 97% of its original size). So, as you can see, preforming a simple task such as image enhancement can be quite complicated procedure and the applications that preform such processing must be lossless – meaning preforming a process like compression or enhancement which allows us to view all the original data when it's needed.

In both issues mentioned above, improvement should be done so that the resulting image is more comprehensible for future analysis than the original one.

Nowadays, when trying to solve a problem using different methods in the field of Digital Image Processing, two main aspects should be considered: first, what is the desirable application? In other words, what kind of problem are we trying to solve? And second, what are the standards for selecting a proper solution to our problem.

So, what is the desirable application?

As we look deeper into image processing industry, we can divide the various applications into the following fields:

  1. Image representation and modeling.
  2. Image enhancement.
  3. Image restoration.
  4. Image analysis.
  5. Image reconstruction.
  6. Image data compression.

What are the standards for selecting a proper solution to our problem?

Given a wide variety of applications available to system designer, how can a choice be made for an application of interest? Clearly, whatever the choice is, it has to be cost effective, nor merely in a financial sense of the word. With this goal in mind, we may identify three important issues that require our attention:

  1. Computational cost.
  2. Performance.
  3. Robustness.

Thus only after inspecting these two important aspects we may start searching for the proper solution.

Whiting an introduction to Digital Image Processing without mentioning even a single formula has been a hard task for me and I hope the general idea has been delivered.

References:

  1. Foundations of digital image processing. Anil K. Jain (ed.) University of California, Davis 1989.
  2. Handbook of Medical Imaging I.N. Bankman (ed.) Academic Press 2000

My special thanks to Molly Ovadia and Constantin Guralnik for their professional consulting.

כתיבת תגובה