Color Project

Color Project
This project requires you to implement and experiment with several programs that manipulate color spaces
in images.
OpenCV support
The programs involve the following color spaces: sRGB, XYZ, Lab, Luv. OpenCV representation of these
color spaces is different from what was discussed in class in the following way: all values are mapped to the
[0 − 255] range and rounded to the nearest integer. Here are the OpenCv conversion routines:
ou tpu t im a ge = c v tC ol o r ( inpu t im a ge , cv2 .FLAG)
FLAG values that are relevant for us are:
COLOR XYZ2BGR, COLOR BGR2XYZ,
COLOR Lab2BGR, COLOR BGR2Lab,
COLOR Luv2BGR, COLOR BGR2Luv.
The project also requires OpenCV code for histogram equalization. Here is how it is used:
o u t p u t g r a y im a g e = cv2 . e q u a l i z e H i s t ( i n p u t g r a y im a g e )
Programs
You are ask to write a total of 9 programs. Each program takes a color image as input and produces a color
image as output. A program operates on a specific color space, and uses 1 out of 3 possible algorithms:
• Linear stretching.
• OpenCV histogram equalization.
• The histogram equalization algorithm described in class.
Please use the following names for your programs:
luv lscl.py luv histeq.py luv classhisteq.py
lab lscl.py lab histeq.py lab classhisteq.py
xyz lscl.py xyz histeq.py xyz classhisteq.py
Each program stretches the illumination component of the image over a specified window. The illumination components are the L part of the Luv and the Lab representations, and the Y part of the XYZ
representation. In each case the illumination component can only have integer values in the [0 − 255] range.
This is not exactly as was described in class. OpenCV makes it easier for us by taking care of the range.
Another point of difference is that the histogram equalization of OpenCV is not exactly as the one discussed
in class.
What you need to implement
Most of your code should use OpenCV routines. You need to implement the two routines that do linear
stretching and class histogram equalization.
Input and output
All programs have the exact same arguments: an input image, window specification, and the name of the
output image. The window is specified in terms of the normalized coordinates w1, h1, w2, h2, where the
window upper left point is (w1, h1), and its lower right point is (w2, h2). For example, w1 = 0, h1 = 0,
w2 = 1, h2 = 1 is the entire image, and w1 = 0.3, h1 = 0.3, w2 = 0.7, h2 = 0.7 is a window in the center of
the image. The provided example program shows how to read the arguments and go over the pixels of the
specified window.
1
Evaluation
Your grade will depend on the following two components: correctness of your programs and your report of
results of experiments. You are asked to experiment with your programs and report the following:
1. Describe strange behavior when colors appear to be changing. If this occurs it is an indication that the
OpenCv code does not handle out of range values properly. You should report it and show an image
where this occurs.
2. Among the 3 lscl programs decide which one is the best and which one is the worst. Show images that
support your conclusion.
3. Among the 6 histeq programs decide which one is the best and which one is the worst. Show images that
support your conclusion.
All experiments should be done with natural images and a window size of at least 50 × 50.
What you need to submit
• Submit the source code of all your programs.
• Submit a report of your experiments.
• Submit at least 3 and at most 10 images to justify the conclusions in your report. Make sure that
these images are unique to your project by keeping them private. They cannot be used in any other
project. (You will lose points if the pictures you submit appear in another project.)
Please notice that it is your responsibility to provide us with all the above information.
You must be present when your project is being evaluated.
Due Date: TBA
2