Image Enhancement

• Process by which we try improve an image so that it looks subjectively better. Better
Means:
q Somehow looks better
q More details are seen
q Contrast is better
q Unwanted flickering has been removed
Enhancement Techniques

• Spatial Domain and Frequency Domain

• The spatial domain technique deals with the manipulation of the pixel values. Hence
can be broadly classified as Point Operation and Mask Operation

• Point Operation : Each pixel is modified by an equation

g(m,n)=T[f(m,n)] T is the operator ( one to one mapping)

• Mask Operation: Each pixel is modified according to values in a small neighborhood (

filtering operation)

Point and Mask Operations

• Point Operation
• Brightness Modification
• Contrast Stretching
• Noise Clipping
• Window Slicing
• Histogram Modelling
• Mask Operation
• Noise Smoothing
• Median Filtering
• Usharp Filtering
• LP/HP/BP filtering

Point Operations
• Zero memory operation
• Given gray level 𝑢 ∈ [0, 𝐿] is mapped into a gray level 𝑣 ∈ [0, 𝐿] according to a
transform
𝑣 = 𝑓(𝑢)

𝑔[𝑥, 𝑦] = 𝑇[𝑓(𝑥, 𝑦)]

or

S=T(r)
Brightness Modification

• Increasing/Decreasing Brightness : Simple method is to add/subtract a constant

value to every pixel of the image.

• If f[m,n] is the original image, then modified image

g[m,n]=f[m,n] + b (increase)
g[m,n]=f[m,n] - b (decrease)

Contrast Stretching

• Transformation T(r) show that it would produce an image of higher contrast than
original by darkening levels below ‘m’ and brightening levels above ‘m’

• Second part of figure produces a 2-level binary image.

• This form of mapping is called thresholding function

• if r1=S1 and r2=S2 it is a linear function ( no change)

• If r1=r2, S1=0 and S2=L-1, it is a thresholding function creates binary image
 • Intermediate values of r1 and r2 produces various degrees of spread in the gray
levels thus affecting the contrast

Gray-level slicing

• Highlights specific range of gray levels in an image

• Fig 1: Display high value for all gray levels in the range of interest and low for all
other gray levels
• Fig 2: Highlights the region of interest but preserves the background and the gray
level tonalities

Some Important Transforms

Image Negative:
• Reverse Scaling of gray level
• 𝑆 =𝐿−1−𝑟
• If L=256, then 𝑆 = 255 − 𝑟
•

LOG transformation

• The general form of the transform is

𝑆 = 𝐶 𝑙𝑜𝑔(|𝑟|+1) where C is the scaling factor.
• This is a non-linear scaling
• Compresses the dynamic range
• Maps a narrow range of low-level gray levels in the input into a wider range in the
output levels
• Used to expand the values of the dark pixels and compress the higher-level values
• Advantage: Compresses the dynamic range of the image with large variation in the
pixel values
• Eg: Displaying 2D Fourier Transform
Power Law Transformation (Gamma Correction)

• This transformation expands the dynamic range by using the relation

𝑆 = 𝐶𝑟 < where C and 𝛾 are constants
• The numerical value of the exponent is called gamma ( Can take both integer as well
as fraction value)
• When 𝛾<1, it maps a narrow range of dark input variables into a wider range of
output values
• Opposite happens when 𝛾>1
• Eg. CRT devices have an intensity to voltage response that is a power function(
exponents varying from 1.8 to 2.5). Hence images tend to be darker
• Gamma correction is done by preprocessing the input image with C=1 and
𝛾=1/2.5=0.4

• The human perception of brightness follows an approximate power function(as

shown below) according to Stevens’ power law for brightness perception (Figure
below)

• If we change input from 0 to 10, the output changes from 0 to 50 (approx.) but
changing input from 240 to 255 does not really change the output value.
• This means that we are more sensitive to changes in dark as compared to bright.

Unlike human perception, camera follows a linear relationship. This means that if
light falling on the camera is increased by 2 times, the output will also increase 2
folds (See Figure below)
Gamma Correction(Display)

• All display devices like computer screen have Intensity to voltage response curve
which is a power function with exponents(Gamma) varying from 1.8 to 2.5.
• This means for any input signal(say from a camera), the output will be transformed
by gamma because of non-linear intensity to voltage relationship of the display
screen.
• This results in images that are darker than intended.

• To correct this, apply gamma correction to the input signal(we know the intensity
and voltage relationship we simply take the complement)
• This gamma is automatically applied by the conversion algorithms like jpeg etc. thus
the image looks normal to us.
• This input cancels out the effects generated by the display and we see the image as
it is.
Histogram Manipulation

• Histogram of an image is the plot of number of occurances of gray levels in the

image against the gray level values
ℎ(𝑟? ) = 𝑛?

Where 𝑟? =kth gray level and 𝑛? =no. of pixels having gray level 𝑟?

• Histogram provides a convenient summary of the intensities of the image

• Does not provide any information about relations amongst pixels

Histogram Plot

• Plots obtained by plotting ℎ(𝑟? ) vs 𝑟? (𝑟? on hor axis and ℎ(𝑟? ) on vert axis
• Plot of dark image clustered around lower gray level
• Plot of bright image clustered around higher gray level
• Low contrast image gives narrow spread of gray level
• High contrast image gives an equal spread of gray level

Histogram Equalization
• Equalization is a process that spread out the gray levels in an image
• Gray levels are equally distributed across their range
• Provides more visually pleasing results
• Let us perform histogram equalization of the image

44444
⎡3 4 5 4 3⎤
⎢ ⎥
f(m,n)=⎢ 3 5 5 5 3 ⎥
⎢3 4 5 4 3 ⎥
⎣4 4 4 4 4⎦
 • Step 1 : Maximum value of gray level (here Maximum value is 5)
• Step 2: Find the minimum no of bits required to represent the value

Eg : If pixel values are between 0-7 , bits required=3 (2I = 8)

If pixel values are between 0-15 , bits required=4 (2K = 16)
If pixel values are between 0-31 , bits required=5 (2M = 32)
If pixel values are between 0-255 , bits required=8 (2N = 256)

• Step 3: Total number of pixels= 25

• Step 4: Find the number of pixels for each gray level
44444
⎡3 4 5 4 3⎤
⎢ ⎥
f(m,n)=⎢ 3 5 5 5 3 ⎥ ;
⎢3 4 5 4 3 ⎥
⎣4 4 4 4 4⎦
• h(0)=0; h(1)=0; h(2)=0; h(3)=6; h(4)=14; h(5)=5; h(6)=0; h(7)=0

Histogram
15

10

0
1 2 3 4 5 6 7

Calculation of Equalized Histogram

Gray Level 0 1 2 3 4 5 6 7

No of Pixels 0 0 0 6 14 5 0 0

Running Sum 0 0 0 6 20 25 25 25

Running
Sum/No. of 0/25 0/25 0/25 6/25 20/25 25/25 25/25 25/25
pixels
Above result X 7 0 0 0 42/25=1.68 140/25=5.6 7 7 7

Round Off 0 0 0 2 6 7 7 7

Original Gray Level 0 1 2 3 4 5 6 7

Final gray Level 0 0 0 2 6 7 7 7

44444 66666
⎡3 4 5 4 3⎤ ⎡2 6 7 6 2⎤
⎢ ⎥ ⎢ ⎥
f(m,n)=⎢ 3 5 5 5 3 ⎥ f(m,n)=⎢ 2 7 7 7 2 ⎥
⎢3 4 5 4 3 ⎥ ⎢2 6 7 6 2 ⎥
⎣4 4 4 4 4⎦ ⎣6 6 6 6 6⎦

Histogram of original and equalized

16
14
12
10
8
6
4
2
0
1 2 3 4 5 6 7