The geometric transformation process needs, as a consequence, to re-assign the radiometric value of each pixel of the new image. In the resampling phase, the original radiometry values are altered.
Resampling techniques are performed by mathematic interpolation procedures.
There are three commonly used resampling algorithms
• Nearest neighbor, the new Digital Number is given by the Digital Number of the pixel having line and row coordinates nearest to the X, Y (real) coordinates obtained by the geometric transformation;
• Bi-linear, the new value is defined by the weighted mean of the values with a surrounding of the four nearest pixels, hence modifying the original radiance value;
• Cubic, the new Digital Number is assigned based on the evaluation of the 16 nearest pixels.
The nearest neighbor has the advantage of assigning in the new image pixel values really observed in the original image. This method is used in classification processing after resampling. Nevertheless, the digital image processing basic rule of performing a classification before any image transformation, when possible, is valid.
Source: Basics of Geomatics
For more Visit: www.agsrt.com