1. This assignment consists of three questions for 20 Marks. Answer all of them.
2. The marks are given alongside the questions.
3. You may use any source on the Internet but should give the reference in your assignment.
4. All programs must be in C, C++ or Python using OpenCV library.
5. If you use any AI Tool, make sure you include the prompt given by you in the assignment. However, remember that you are being very bad to the environment! You can power 100 high-end apartments in Hyderabad for a month with the power AI tools consume in one hour.

• Your assignment MUST BE DONE in a Jupyter notebook.
• The name of the notebook must be
  <your roll number>_assignment_III with the roll number being in uppercase letters.
• Every question must be done in the following manner in the notebook
  • Create a markdown cell with a copy of the question.
  • Create as many code cells as needed to implement the algorithms asked.
  • Every cell MUST HAVE a comment line as the first line(s) that say(s) what the cell contains.
  • There can be one or more markdown cells after your code to present your findings and discussions.
  • In case you are implementing your programs in C/C++, use the magic commands such as %⁠%file and %⁠%bash to show execution in the notebook.
• The final notebook must be submitted by email with the subject CIP Assignment III to chakravarthybhagvati@uohyd.ac.in by the due date, Friday, 08 May 2026, 9:00 PM.

1. (10 Marks)Write a program that converts a given spectrum into an RGB colour. It should have the following prototype in Python
   spectrum2rgb(<input spectrum>)
   where the spectrum is a 61x1 vector with the spectral values in 5 nm intervals from 400 nm to 700 nm (both inclusive). In C/C++, the same function has the prototype
   int spectrum2rgb(float <input spectrum[61]>)
   Now, write a program that converts an RGB triplet into a spectrum using the pseudo-inverse method that we did in class. The prototypes must be as follows:
   Python: rgb2spectrum(<rgb colour>)
   C/C++: int rgb2spectrum(int <rgb[3]>, float <spectrum[61]>)
   Compare the original spectrum with the one that you computed using the pseudo-inverse. This code requires a lot of tweaking the numbers to get RGB values and recover similar spectra. Try for 10 different Munsell Colours and write down your own observations on similarities and dissimilarities between the original and computed spectra. Hint: Read the various sources on the Internet for converting spectra into RGB colours.
2. (6 Marks) Implement the median-cut quantisation algorithm. Use it reduce an image to 256, 128, 64, 32 and 16 colours respectively. You can try any of the images given in the previous assignment. Function prototypes:
   Python: median_cut(<input image>, <no. of colours>) The function returns the quantised image.
   C/C++: int median_cut(<input image>, <quantised image>, int <no. of colours>)
3. (4 Marks) Briefly explain the gif image format and state precisely why it requires quantisation.

Data for the assignment

1. Munsell Spectra Dataset (CSV format)
2. CIE Colour Matching Functions (CSV format)
3. D65 Spectral Data (CSV format)
   You may need to use it for normalising Munsell Spectra and get the correct RGB values. Of course, its need depends on the method you use; some methods on the Internet build it into their algorithms implicitly.

1. This assignment consists of three questions for 20 Marks. Answer all of them.
2. The marks are given alongside the questions.
3. You may use any source on the Internet but should give the reference in your assignment.
4. All programs must be in C, C++ or Python using OpenCV library.
5. If you use any AI Tool, make sure you include the prompt given by you in the assignment. However, remember that you are being very bad to the environment! You can power 100 high-end apartments in Hyderabad for a month with the power AI tools consume in one hour.

• Your assignment MUST BE DONE in a Jupyter notebook.
• The name of the notebook must be
  <your roll number>_assignment_II with the roll number being in uppercase letters.
• Every question must be done in the following manner in the notebook
  • Create a markdown cell with a copy of the question.
  • Create as many code cells as needed to implement the algorithms asked.
  • Every cell MUST HAVE a comment line as the first line(s) that say(s) what the cell contains.
  • There can be one or more markdown cells after your code to present your findings and discussions.
  • In case you are implementing your programs in C/C++, use the magic commands such as %⁠%file and %⁠%bash to show execution in the notebook.
• The final notebook must be submitted by email with the subject CIP Assignment II to chakravarthybhagvati@uohyd.ac.in by the due date, Wednesday, 08 April 2026, 9:00 PM.

1. (7 Marks) Write a program to implement dispersed-dot dithering. The program MUST be organised into the following function
   Python:  
   apply_dither_mask(<input image>, <dither_mask>, <colour_mask>)
   The return value is the output image.
   C/C++:
   int apply_dither_mask(<input image>, <output image>, <dither_mask>, <colour_mask>, int mrows, int mcols)
mrows, mcols are the number of rows and columns respectively in the dither mask. The return value is \(0\) on success and any other value for errors.
   Experiment with \(4\times4\) and \(5\times5\) dither masks. You may use flower image and fall colours image in your experiments.

   Discuss the following based on your results

   • How did you spread the threshold values and the colour component positions?
   • Did you allocate different number of thresholds to different components? Did you think it made any difference to quality of the output?
   • Did the image resolution have an effect on the quality of the output?

2. (6 Marks) Implement Floyd-Steinberg error diffusion algorithm for colour images. Use any method discussed either in the class or you find on the Internet. Write the method clearly as an algorithm and then show its working on the images from the previous problem. Compare your results with dithering; discuss times taken and the quality of the outputs.
3. (7 Marks) Use any of the Bayer Colour Filter arrays (get them from the Internet) and implement an algorithm that simulates image formation using your selected Bayer CFA. Remember that you get a grayscale image.

   Implement an appropriate demosaicking algorithm to restore a full colour image from the grayscale Bayer CFA image. Measure the difference between the original image and the demosaicked image by subtracting one from the other and taking the absolute values. Remember that an image is like a matrix and we know how to subtract matrices, don't we?! What is the mean error per pixel? Compute it as the sum of the errors / number of pixels in the image.

   You can use the same images as for the above problems.

There are a few more images here for you to play with:
• Cat (high-detail)
• Door Knocker (shiny detail)
• Pink Flower
• Yellow Flower
• Rainbow (very challenging for dithering)

---

end of assignment ii

1. This assignment consists of three questions for 20 Marks. Answer all of them.
2. The marks are given alongside the questions.
3. You may use any source on the Internet but should give the reference in your assignment.
4. If you use any AI Tool, make sure you include, in the assignment, the prompt given by you. However, remember that you are being very bad to the environment! You can power 100 high-end apartments in Hyderabad for a month with the power AI tools consume in one hour.

1. (8 Marks) Use the book, Digital Image Processing, to find the equations for converting RGB to L*a*b* space and vice-versa. Write a Python module or C functions on your own to implement the conversions. Hint: You may want to use the openCV library in both Python and C languages.
   Use your code to find the pair-wise distances between the colours: C1 = (64, 192, 96), C2 = (168, 255, 128) and C3 = (64, 66, 96) in RGB and L*a*b* spaces. Which distances appear to match your perception of the three colours?
2. (4 Marks) Consider the colour C = (0, 32, 128). Convert it into C^hsv using Python colorsys/openCV libraries. Multiply the intensity value by 1.5 and then convert the colour back into RGB space. Multiply the original C in RGB space by 1.5. In other words, multiply each component by 1.5. Are the two colours the same? Explain your answers.
3. (8 Marks) Download the Colour Matching Matrix in csv format. This contains the CIE colour matching functions from 400 nm to 700 nm in 5 nm intervals (i.e. 61 rows and 3 columns).
   Using the data, write programs for two functions: to create a spectrum as follows create_gauss_spectrum(center, bandwidth, energy) The parameter center is the wavelength, expressed in nanometres between 400 and 700nm, bandwidth is the width of the spectrum and energy is the total area under the spectrum. The spectrum is a gaussian function centred at the wavelength with its standard deviation given by the bandwidth divided by 6. The gaussian function is defined as $$\mbox{gauss}(x, m, w, A) = \frac{A}{\sqrt{2\pi}w}e^{-\frac{(x-m)^2}{2w^2}}~~~ \mbox{(Corrected,}~`-'~\mbox{missing in exponent earlier)}$$ In this equation, \(m = \mbox{center}, w = \frac{\mbox{bandwidth}}{6}, A = \mbox{energy}\).
   Important: Remember to write the function such that the spectrum is also in steps of 5 nm from 400 nm to 700 nm.
   The second function is to convert a spectrum into a colour spect2rgb(spect) where spect is the input spectrum and the return value from the function is an RGB triplet. If you are using a C program, use extra parameter(s) to get the return value as it is not easy to return an array.
   You also need the XYZ to RGB conversion matrix (3x3). Download or copy the XYZ to sRGB matrix for D65 illuminants from any source on the WWW.
   Use your functions to create three gaussian spectra with different sets of parameters and convert them into RGB colours. Are the colours what you expect? Explain briefly.

Chapter 6: Colour Image Processing
Sections covered

• Basics of Full Color Image Processing
• Color Transformations
• Color Image Smoothing and Sharpening

Link for Vector Image Processing (also available as Item IV in the Syllabus section).
Also, Material from class

Minor-I Answers