Digital Image Processing Techniques


See how to operate the DIP techniques in SPRING

Techniques

Introduction to Image Processing

The digital analysis of data, more specifically of remote sensing digital imagery from orbital platforms, has seen in the last 25 years a great deal of developments of techniques aimed at the analysis of multidimensional data, acquired by a number of different sensors. These techniques have received the name of digital image processing.

By  digital image processing we mean the manipulation of an image by a computer in such a way that both the input and the output are images. By comparison, in the pattern recognition discipline the input is an image while the output consists of a classification or a description of this image. On the other hand, the area of computer graphics consists of the generation of images starting from a description of them.

The objective of using digital image processing is to enhance the visual aspect of certain structural features in the image for the human analyst and also to provide other subsidies to its interpretation, including the generation of products that could later undergo further processing.

The digital image processing area has attracted lots of interest in the last two decades. The evolution of the digital computing technology as well as the development of new algorithms for the treatment of bidimensional signals are creating an increasing number of applications.

As a result of such evolution, the digital image processing technology is widening its domain, that includes many different areas like, for example: analysis of natural resources via satellite imagery; digital transmission of television or facsimile signals; biomedical image analysis, including the automatic cell count and the exam of chromosomes; analysis of metallography images and vegetal fibre images; medical imaging by ultrasound, nuclear radiation, or computed topography techniques; application in industrial automation involving the use of robotic vision, etc...

The use of multispectral images obtained by satellites like Landsat, SPOT, ERS1, NOAA, and others, have proved to be a powerful technique for many applications in the research of natural resources. The acquisition of spectral information obtained by the systems in different parts of the electromagnetic spectrum, aimed at the identification and discrimination of targets, depend basically on the quality of the data representation contained in the images. The digital processing techniques available in SPRING are better suited to the images generated by multispectral sensors.

The techniques of digital image processing (DIP), besides allowing the analysis of a scene in various regions of the electromagnetic spectrum, also allow the integration of many different data types, duly registered.

The digital image processing can be divided in three different steps: pre-processing, enhancement, and classification. The pre-processing refers to the initial processing of the raw data for the radiometric calibration of the image, the correction of geometric distortions and the elimination of noise. The most common enhancement techniques in DIP are: contrast enhancement, filtering, arithmetic operations, IHS and principal components transformation. As to the classification techniques we have: supervised classification (by pixel), and unsupervised classification (by regions).

NOTE: The user may choose not to use the classification algorithms, since he can decide to perform the direct interpretation of an enhanced image.

As we will see the DIP techniques are always over the gray levels or digital numbers (DN) attributed to each pixel in the image. Depending on the technique used the user will work with one image only (one band or IL) or with many images, this last one known as multispectral techniques, because it deals with many images of the same region in different regions of the electromagnetic spectrum.

See following about each of these techniques:


Digital Image Processing Techniques


See also:
About Radar Image Processing
How to register an image ?