5
\$\begingroup\$

A colleague mentioned a course in digital signal processing where he remembered the expression "normalizing the curve" or "normalizing the image" to make it as accurate as possible and get better image resolution and small details. He drew a histogram of a picture with three sharp peaks and another histogram after normalization with the peaks in the same locations but with larger spread.

The colleague could remember nothing more. I could not find a reference for those expressions. It sounds to me like gamma correction with RGB photos or High Dynamic Range in RAW photos.

Is that right, or is that a different technique?

Update: The colleague was processing photos in Matlab, in case that's relevant.

Improve this question
\$\endgroup\$
5
  • 1
    \$\begingroup\$ Sounds like gamma to me (setting white/black points). \$\endgroup\$
    – Steven Kersting
    Commented Dec 8, 2019 at 15:40
  • \$\begingroup\$ Anyone else think of statistical normalization? \$\endgroup\$
    – xiota
    Commented Dec 8, 2019 at 19:00
  • \$\begingroup\$ @Tetsujin With audio signals, if you normalize values, you raise or lower the actual values so that the peak (or average or floor is adjusted to a specific level and the rest is raised or lowered by an equal or proportional amount). \$\endgroup\$
    – Michael C
    Commented Dec 9, 2019 at 3:49
  • 1
    \$\begingroup\$ I know that Paint Shop Pro has or had a function called Histogram Equalization that would enhance or deemphasize different brightness bands in an attempt to flatten the histogram. \$\endgroup\$
    – Mark Ransom
    Commented Dec 10, 2019 at 16:50
  • \$\begingroup\$ Today I learned about color mapping and I get the impression "normalizing the curve" could be transferring the color of one or more reference histograms onto the image. \$\endgroup\$
    – emonigma
    Commented Dec 18, 2019 at 23:04

1 Answer 1

Reset to default
4
\$\begingroup\$

What your colleague meant, based on your description, was: Setting the white and black point of the image.

On an 8-bit grayscale image, you have 256 different values ranging from total black (0) to total white (255) and various shades of gray (values 1-254).

An image with content only in the ranges between e.g. 0 and 120 (under exposure) might feature "three sharp peaks". Setting the white point to 120 will make the peaks in the histogram have 'larger spread'. However, the absolute location of the peaks must change in this process (unlike what your colleague has drawn).

Widening the peaks without change of peak location is definitely not what is meant by 'normalization' - still considering, that this term is used in different places with different meanings.

Improve this answer
\$\endgroup\$
1
  • \$\begingroup\$ The colleague didn't label the axes but I assume the leftmost was near 0 and the rightmost near 255. What is the name of widening the peaks without changing location? \$\endgroup\$
    – emonigma
    Commented Dec 15, 2019 at 22:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.