Image Processing

• We load an image of the Mars moon Phobos:

      im1 = imread('phobos.png');

• and display it:

      imagesc(im1);

• Since the image was taken with a black-and-white CCD, we change the colormap:

      colormap(gray);

• Im1 contains values from 0 to 255, representing arbitrary colors in a color table. But in our case these are just lightness values, so we change the representation to a usual array, containing double values between 0 and 1.

      im2 = double(im1) / 255;

• We control our variables and find that the new image consumes 8x more memory:

      whos
       Name      Size         Bytes  Class
     
       im1     292x510       148920  uint8 array
       im2     292x510      1191360  double array
     
     Grand total is 297840 elements using 1340280 bytes

• Obviously the edges contain bad data, so we cut them off:

      im3 = im2(3:289, 5:509);
      imagesc(im3);

• Some strange values are due to defect CCD pixels. To find them, we create a smoothed image where every pixel is replaced by the middle value of its neighbours ("median filter") and look for large differences between the original and the smooth image:

      smooth = medfilt2(im3);
      diffs = abs(im3 -smooth);
      badspots = diffs > 0.2;
      imagesc(badspots);

• We replace the values at the bad spots with the corresponding smoothed values:

      im3(find(badspots)) = smooth(find(badspots));
      imagesc(im3);

• Finally, we save our cleaned image:

      imwrite(im3, 'clean.png');