1. Image Processing – Contrast Enhancement
Jack Tompkins
Department of Computer Science

2. What is Contrast Enhancement
Original Image with low contrast Enhanced Image

3. Luminance
While contrast enhancement can be accomplished across r, g, and b bands, often the image is first converted to gray scale.
When translating a color image to black and white (mode “L”), use the ITU-R luma transform:
L = R * 299/ G * 587/ B * 114/1000

4. Contrast
Low contrast images with neighbors clustered on the low end of the intensity scale, closer to 0, are called low-key.
Low contrast images that reside on the high end of the intensity scale, closer to 255, are called high-key.
Normal contrast images are equalized.
High contrast images will exhibit a full range of tones from black to white, with dark shadows and bright highlights.

5. How to Enhance Contrast
Neighboring pixels with similar intensity are hard to distinguish. Low Contrast -> image values concentrated in a narrow range.
Contrast enhancement -> distribute the pixel intensities across a broader range while maintaining relative brightness

6. Visualizing Pixel Counts Histograms
A histogram is a display of statistical information representing the frequency of data items in successive numerical intervals of equal size.
Histogram of a monochrome image with 256 possible gray levels or intensities ranging from 0 through 255.
Each bin in the histogram represents a probability. P(i) = ni / n, where ni is the number of pixels with intensity i, n is the total number of pixels.

7. Pixel Counts - Histograms

8. Visualizing Pixel Counts Histograms
histograms.xslx
Bw1.jpg
Luther Bell
Lena
Oira.jpg

9. By What Technique?
Use a function, g, to generate a new image B from image A: B(x, y) = g(A(x, y)), x = 0,…, n-1, y = 0,…, m-1
The function, g, maps pixel intensities for each pixel in image A to a new intensity for that pixel in image B.
Monotonically non-decreasing, (relative intensity relations remain)
A short chapter excerpt with examples: TT92

10. Linear Transform Uniform linear transformation
Say intensities are concentrated from 0 to 64, and we wish to map these to 0 to 255
g(f) = m f + b, where m = (t1 – t0)/(s1 - s0), b = t1 – m s1 g(f) = 255/64 * f – 255/64 * 64 = 255/64 f see histograms.xslx
Easily adapted to piecewise transformations using multiple slope segments over sequential groups
Can be fully automated

11. Histogram Equalization
Digital Image Processing – Gonzalez & Woods, p 93
The probability of occurrence of gray level rk in an image is approximated by
p(rk) = nk / n k = 0, 1, 2, …, L-1, where nk is the number of pixels with gray level rk, n is the number of pixels, and L is the total number of possible gray levels in the image (typically 256)
A plot of pr(rk) is called a histogram, efficiently stored in an array with indices

12. Histogram Equalization
Digital Image Processing – Gonzalez & Woods, p 93
A processed output image is obtained by mapping each pixel with level rk in the input image into a corresponding pixel with level sk in the output image
sk = (L-1) 𝑗=0 𝑘 𝑝 (rj) = (L-1) 𝑗=0 𝑘 𝑛 𝑗 𝑛 = ((L-1)/n) 𝑗=0 𝑘 𝑛𝑗
Efficiently stored in an array with indices a look up table
Histogram Equalization

13. Additional Techniques
Image Contrast Enhancement Methods
Log transform: g(f) = c log(1 + f)
Power transform: g(f) = c f r, 0 < r < 1

14. Saving multiple image histograms to a single CSV file

15. Data gathering

16. Scanning in the Image Files