CMPE 464  Project 4:

Image Enhancement - Due: December 19th

 

 

Given:   Three images, lena.bmp, lenag.bmp, lenai.bmp

           

Required: Write programs to implement the following techniques:

1.    Apply linear filtering to the noisy images, lenag.raw and lenai.raw. Use two different filters: A square-shaped averaging filter and a gaussian filter. The filter window size user be user-selectable. Note that in each case, the filter kernel should sum to one. That is, the output pixel value should be divided by the sum of the values in the kernel. At the edges, assume that the image extends outside. Assume that the values outside are the same as the closest pixels inside. (This is equivalent to appending a number of rows and columns at the top, bottom, left and right).

2.    Apply nonlinear filtering to the noisy images, lenag.raw and lenai.raw. Use two different filters: A square-shaped standard median filter and a square shaped alpha trimmed mean filter. The filter window size and alpha for the alpha trimmed mean should be user-selectable.

In each case, find the mean sqared error(MSE) and tabulate MSE values. MSE is defined as the average squared error between corresponding original image pixels and filtered noisy images.

 

Take printouts of the outputs in each case and discusss:

·        The effect of linear filtering on two types of noisy images

·        Averaging filter vs. gaussian filter – which one do you prefer?

·        The effect of linear filter size

·        The effect of nonlinear filtering on two types of noisy images

·        Median vs. weighted median – do you see advantages with weighted median? Which weights are best?

·        The effect of filter size