108
CHAPTER 7. NOISE
Figure 7.13: Using Wiener filtering to remove Gaussian noise with low variance
4. Pratt [16] has proposed a “pseudo-median� filter, in order to overcome some of the speed disadvantages of the median filter. For example, given a five element sequence , its pseudo-median is defined as
psmed
So for a sequence of length 5, we take the maxima and minima of all subsequences of length , we take the maxima and three. In general, for an odd-length sequence of length minima of all subsequences of length . We can apply the pseudo-median to neighbourhoods of an image, or cross-shaped neighbourhoods containing 5 pixels, or any other neighbourhood with an odd number of pixels. Apply the pseudo-median to the images in question 1, using neighbourhoods of each pixel. 5. Write a Matlab function to implement the pseudo-median, and apply it to the images above with the nlfilter function. Does it produce a good result? 6. Produce a grey subimage of the colour image flowers.tif by >> f=imread(’flowers.tif’); >> fg=rgb2gray(f); >> f=im2uint8(f(30:285,60:315)); Add 5% salt & pepper noise to the image. Attempt to remove the noise with (a) average filtering, (b) median filtering, (c) the outlier method,