Scribd
Upload
69 views

Histograms: CS 4640: Image Processing Basics

This document discusses histograms and how they relate to image properties. It defines a histogram as a function that gives the frequency of intensities that occur in an image. Histograms can show properties like exposure, brightness, contrast, and dynamic range. Simple point operations like addition and multiplication affect the histogram. Binned histograms divide the intensity range into bins. Histograms can be interpreted as probabilities by relating pixel counts to total pixels. This allows defining sample spaces, events, and expectations for images.

Uploaded by

Muhammad Zaka Ud Din
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
69 views

Histograms: CS 4640: Image Processing Basics

This document discusses histograms and how they relate to image properties. It defines a histogram as a function that gives the frequency of intensities that occur in an image. Histograms can show properties like exposure, brightness, contrast, and dynamic range. Simple point operations like addition and multiplication affect the histogram. Binned histograms divide the intensity range into bins. Histograms can be interpreted as probabilities by relating pixel counts to total pixels. This allows defining sample spaces, events, and expectations for images.

Uploaded by

Muhammad Zaka Ud Din
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 14

Histograms

CS 4640: Image Processing Basics

January 19, 2012


Histograms

Definition
A histogram is a function that gives the frequency of
intensities that occur in an image.

So, given an image I : [0, K 1], the histogram


for I is the function

h(i) = card{(u, v)| I(u, v) = i}.


In other words

h(i) = the number of pixels with intensity i


Histograms
How Image Properties Show Up in
Histograms (ImageJ Examples)

I Over or Under Exposure


I Brightness
I Contrast
I Dynamic Range
How Simple Point Operations Affect
Histograms (ImageJ Examples)

I Addition
I Multiplication
I Exp and Log
I Intensity Windowing (Contrast Modification)
Histograms For More Than 8 Bits

Definition
A binned histogram is a function that gives the
frequency of image intensities that fall into small
intervals, or bins.

So, given the image I : [0, K 1], the binned


histogram for I is the function

h(i) = card{(u, v)| ai I(u, v) < ai+1 },

where 0 = a0 < a1 < . . . < aB = K .

Typically, we choose equally spaced bins: ai = iK/B


Images And Probability

Histograms can be interpreted as probabilities.

Question: If I pick a pixel from an image at random,


what is the probability that the pixel has intensity i?
# pixels with value i
Answer: P(I(u, v) = i) = total # pixels

Or, in terms of the histogram h:

P(I(u, v) = i) = h(i)/wh
Sample Space

Definition
A sample space is a pair (S, P), where S is a set of
possible outcomes, and P : S R is a function that
satisfies P(s) 0 for all s S and
X
P(s) = 1.
sS
Sample Space for Images

Given an image I(u, v), we can define the S as the set of


possible outcomes, i.e., possible intensities of a pixel.
So, S = [0, K 1].

The probability function P is defined via the histogram:

P(i) = h(i)/wh
Events

Definition
Let (S, P) be a sample space. An event is a subset
A S. The probability of an event is
X
P(A) = P(a)
aA
Some Possible Events for an Image Pixel
I A = The pixel has value i

P(A) = P(i) = h(i)/wh

I A = The pixel has value < t


t1
X
P(A) = P(a)
a=0

I A = The pixel value is in the interval [70, 93]


93
X
P(A) = P(a)
a=70
Random Variables

Definition
A random variable is a function defined on a probability
space. In other words, if (S, P) is a sample space, then a
random variable is a function X : S V for some set V .

I A random variable is neither random nor a variable!


I We will only be worried about the identity function
X(i) = i.
I I will use the notation I to denote an image, or the
random variable associated with it.
Expectation

Definition
Let X be a real-valued random variable defined on a
sample space (S, P). The expectation, or expected
value, of X is
X
E(X) = X(s) P(s)
sS
Expectation of Pixel Values

Keep in mind: P(i) = h(i)/wh

Mean of an image:

K1
X
= E(I) = i P(i)
i=0

You might also like