A New Selective Filter for Improving the Performance of Geotomographic Algorithms

The convergence of iterative geotomographic algorithms can be improved by applying an appropriate smoothing filter between some steps of the algorithms. A new filter for accomplishing such a task is presented. The filtering process consists of passing a twodimensional window throughout the image. The window is divided into four groups of pixels, and a block is formed by the union of the two groups that most likely belong to the same feature of the image as the pixel whose amplitude is being calculated. The new amplitude of this pixel is the average of the amplitudes of the other pixels in the block. Whenever a new amplitude is calculated, it is immediately assigned to the respective pixel. The filter smooths within the features without blurring their boundaries. Results of a computer simulation illustrate the ability of the new filter to improve the convergence behavior of tomographic algorithms. When the filter was employed, an electrical conductivity geotomograph required fewer iterations to be reconstructed and presented much fewer artifacts. Since the filter automatically adapts to the characteristics of any image, it may also be used to process images created by geophysical methods other than geotomography.