Hand in a short report on your work, which should be two written sides of A4 at most, plus any pictures and plots showing your results and a print out of your code. Write a short description of the methods you used for each part together with any conclusions you have drawn from your experiments. Make sure it is clear what commands you used to generate each of the pictures you include.

The hand-in deadline for this coursework is 12:00 noon on 14th January. Hand your report into the CS departmental office. Make sure you attach a completed coursework cover sheet to your work.

In general, we seek color image descriptions that characterize the different objects' materials in a way that is independent of a) the geometry of the illuminated surface (slides, p.28), b) the illumination color (p.49), or sometimes c) both. In this case:

- Define regions* corresponding to each of the three objects in the image above. Show the mask or outline of the regions. Generate plots that show the occupancy of the RGB cube (i.e. which colors appear in the object's image region) for each object. A nice way to do this is to plot each object using a different color. Look up the matlab function plot3 and see a sample here to see how to generate a simple occupancy plot. Comment on how the color distributions vary among objects.
- Compute the mean and covariance of the color distribution for
each object under this illumination. A simple measure of separation of
two Gaussian distributions with means
*m*_{1}and*m*_{2}and covariances**C**_{1}and**C**_{2}is min{*d*_{1}(*m*_{2}),*d*_{2}(*m*_{1})}, where*d*_{i}(x) = (x-*m*_{i})^{T}**C**_{i}^{-1}(x-*m*_{i}). Construct a table of separations between the three objects' color distributions. Comment (only) on how reliably you expect a computer vision system that uses the unprocessed color information alone to recognize and distinguish between these different object surfaces in unknown illumination. - Construct occupancy plots after correction for surface geometry, i.e. in "normalized color" space. Show the image in this new color space. Recompute the mean and covariance matrices in the normalized color space. Recompute the table of separations between each pair of distributions and comment on whether you would expect any improvement in object surface recognition using normalized color.
- Construct occupancy plots for the image above after correction for illumination color only. Show the image in this new color space.

(Note, unlike the previous courseworks, these 50% of the marks should require less effort than usual)

- Correct the below image for illuminated surface geometry.
- Correct the below image for illumination color.
- What color (red, green or blue) is the pipe-shape in this second image? It may seem bright gray, but actually has some saturation. Correct for anything else you need, and show your work, justifying your answer. There are several acceptable solutions & explanations.

*= You can use imfreehand(), for example:

imshow(I);

h = imfreehand

IbwMask = createMask(h);