Histogram Equalization

HistogramEqualization.py

from PIL import Image

# Converts image to grayscale
inputImage = Image.open("source.png").convert('L')   
outputImage = Image.new('L', inputImage.size)

width, height = inputImage.size

frequency = {}

# There's simpler ways of writing the following code, but I'm prioritizng clarity over brevity
# Fill the frequency count with 0s
for intensity in range(256):
	frequency[intensity] = 0

# Count how frequently an intensity level appears in the input image
for x in range(width):
    for y in range(height):
        intensity = inputImage.getpixel((x,y))

        if (intensity in frequency):
        	frequency[intensity] += 1
        else:
        	frequency[intensity] = 1

intensityFrequencies = [frequency[key] for key in sorted(frequency.keys())]

cummulativeSum = 0
index = 0

cdf = []
for intensityFrequency in intensityFrequencies:
	cummulativeSum += intensityFrequency
	cdf.append(cummulativeSum)
	index += 1

# Now we need to normalize the CDF to fit between a range of [0, 255] 
# We use the values in the CDF to change the intensity of pixels in the original image so we need our intensity values to be normalized to [0,255]

minCDF = cdf[0]
maxCDF = cdf[-1]

normalizedCDF = []

for value in cdf:
	# Reminder: We need to multiply by 255 as normalization gives us a value between 0 and 1
	normalizedCDF.append(((value - minCDF) * 255) / (maxCDF - minCDF))

print(normalizedCDF)
print(len(normalizedCDF))

for x in range(width):
    for y in range(height):
        intensity = inputImage.getpixel((x,y))
        outputPixel = int(normalizedCDF[intensity])
        outputImage.putpixel((x,y), outputPixel)

outputImage.save('histogramEqualization.jpg')