Relative Efficiency of Adaptive Filters in Suppressing Noise And Preserving Sharpness in Multi-Spectral Images

Shivaji Prasad, Department of Geography, Frostburg State University, sprasad@frostburg.edu

Influx of suspended sediments and other soluble materials in coastal waters considerably influence light reflectance characteristics of these waters; and make it possible to separate into clean and turbid classes. However, random noise signals introduced in digital imaging make it difficult to separate turbid waters further into different turbidity classes. In order to make these classes separable, there is a need to sharpen the boundaries between turbidity classes. Since, adaptive filters have shown promise in removing speckles and preserving image sharpness in radar images, it was speculated that these filters might help in sharpening boundaries making class separation possible.

In view of this, the objective of this research was to examine the efficiency of several adaptive filters in terms of preserving sharpness of boundaries between waters of varying turbidity levels. In this study, subsets of a TM scene of turbid waters present along the coast of Northern Australia were prepared, and subjected to adaptive filtering. Several of the adaptive filters namely Lee, Frost, Kuan, Gamma and Local Sigma were applied. The results indicated that, in general, Local Sigma filter produced sharper images of turbid waters and improved interpretability. Relative efficiencies of these filters along with their comparisons were included in this study.