Description

Grading: 25% for implementation and 75% for written report

Mon Tue Wed Thur Fri
James Fishbaugh jf146@nyu.ed 4-5 PM 4-5 PM
Michael Lally mfl340@nyu.ed 12-1 PM 10-11 AM
Akshat Khare ak7674@nyu.edu 2-3 PM 10-11 AM

Please read the instructions carefully. Note: we will not be running code. Rather, we will check your code to make sure your implementation is your own, and it matches your results. Your grade is primarily based on your written report. This means going beyond just showing results, but also describing them. You should produce a standalone lab report, describing results in enough detail for someone else (outside of class) to follow. Please submit a single PDF/HTML with all code included as an appendix.

A) Programming Questions

The purpose of this project is to get familiar with the coding environment for the rest of the course, reading/writing images, displaying images, and making graphs. You will implement histogram equalization.

Hint for implementation: You can assume an intensity range [0-255] and hardcode the size of the histogram to be 256. You can make sure you have an image of type uint8 with min 0 and max 255 with the python code:

# Convert the image to type uint8 and scale intensity values to the range 0-255
# *Note* numpy min/max flatten the 2D array, so you obtain the min/max of the entire image im_uint8 = ((im – np.min(im)) * (1/(np.max(im) – np.min(im)) * 255)).astype(‘uint8’)

A1) Histogram Equalization

Implement the following functions:

def create_pdf(im_in):
# Create normalized intensity histogram from an input image return pdf

def create_cdf(pdf):
# Create the cumulative distribution function from an input pdf return cdf

def histogram_equalization(im_in):
pdf = create_pdf(im_in) # Your previously implemented function cdf = create_cdf(pdf) # Your previously implemented function # Create a histogram equalized image using your computed cdf return equalized_im

Write up a report including the following:

• Include a brief introduction and description of how histogram equalization works.

• Show crowd.png before and after histogram equalization, and the corresponding histograms (PDFs).

• Discuss how the image and histogram have changed, and connect it back to your description in 1).

• Show the cumulative distribution function before and after histogram equalization on the same figure. Describe what you see. Explain the shape of each CDF and relate it back to image contrast and intensity histogram shape.

• Reapply the histogram equalization procedure on the corrected image. Show and discuss the results.

• Apply histogram equalization to another low contrast image (greyscale). Show and discuss the results.

• Histogram equalization is a global solution, modifying contrast with respect to the CDF of the entire image. Imagine you split the image into smaller regions (e.g. patches of 50×50 pixels) and apply histogram equalization to each local patch independently. In your own words, describe advantages and disadvantages of this strategy. There are no right or wrong answers here, the idea is to think critically about the method and answer honestly in your own words.

• Include all code as an appendix, i.e. copy and paste all your code at the end of your report.