Bilateral filtering is a non-linear filtering technique introduced by Tomasi et al. [1]. It extends the concept of Gaussian smoothing by weighting the filter coefficients with their corresponding relative pixel intensities. Pixels that are very different in intensity from the central pixel are weighted less even though they may be in close proximity to the central pixel. This is effectively a convolution with a non-linear Gaussian filter, with weights based on pixel intensities. This is applied as two Gaussian filters at a localized pixel neighborhood , one in the spatial domain, named the

Let

where t is the scale parameter.

For the central pixel of the neighborhood a

Pixels close to the central pixel a

We apply the bilateral filter above to our original image before the laplacian edge detection to smooth out the noise while maintaining edges. We can then enhance the edges without amplifying noise. Fig 9 shows the application of the bilateral filter on same noisy image of Lena from the previous section . Comparing the performance with vs. without bilateral filtering, the improvement both in noise reduction and edge-preservation should be apparent.

Fig 9: Bilateral Filtering for noiseless edge preservation.

It is interesting to note that we explored application of the same bilateral filter after scaling (adjusted corrected for new domain scale) rather than before as shown in Fig 10. The results were very comparable (Fig. 9). It is computationally cheaper hence preferable to perform the bilateral filtering prior to scaling.

<Previous Page> <Next Page>