TA feedback

Digital Photography, spring 2012

Exercise 1, Matlab introduction


Exercise 2, Image formation

  • Linear transform between two color spaces: [Rout Gout Bout]’ = T*[Rin Gin Bin]’
    This means that the 1st row in T defines the weights for the input R, G, B values to determine the output R value. 2nd row for output G and 3rd row for output B.
  • The multispectral image with the NBS basketball and the tiled wall in the backgrouns has a 4 pixel wide black bar on the left hand side. Better remove to avoid confusions.


Exercise 3, White balance

  • Exercise 5.4: Explain how to apply a Gaussian filter to an image. Note: This is actually done in Exercise Session 4, where it is no longer useful.
  • The name “Weighted Grey World” is confusing (and lead to confusion), because it actually is an unweighted Grey World.
  • The “Weighted Grey World” algorithm should be modified since at the resolution of the provided images, every bin gets at least one pixel, and hence this algorithm does nothing at all. We could either define that we only want to count the bins if there is a certain number of pixels based on image statistics, or downsample the image.
  • The demo image for the MaxRGB exercise needs to be changed, as the max values of all three color channels are very close to 1.0 and so the effect is almost not visible.


Digital Photography Spring 2013


Exercsie 1, MATLAB Introduction.

– Some introduction to plots and also how to use functions in MATLAB need to be added.


Exercise 2, Image Formation

– Image normalization has to be explained more clearly.

– The T matrix used to transform the camera color space to linear sRGB, does not preserve the white point. It needs to be changed.

– True color display for Cr and Cb channels seems to be difficult to understand for students at this stage of the course. It’s better to remove this part.

– Correct sRGB to YCbCr conversion includes both linear tranformation ans shifting the channels. Up to this point, we always asked students to implement only the linear part. It is better to either addd the shifting or ask them to use MATLAB rgb2ycbcr function.