COMPUTER VISION (Module-1 Chapter-3)

Image Processing
Dr. Ramesh Wadawadagi
Associate Professor
Department of CSE
SVIT, Bengaluru-560064
[email protected]
Visualizing Image Data (Grey scale)
Visualizing Image Data (Color image)
Point operators
● The simplest kinds of image processing
transforms are point operators, where each
output pixel’s value depends on only the
corresponding input pixel value.
● Examples of such operators include
brightness and contrast adjustments as well
as color correction and transformations.
● Point operators are also called as pixel
transforms.
Pixel transforms
● A general image processing operator is a function
that takes an input image and produces an output
image.
● In the continuous domain, this can be denoted as:

g(x) = h(f(x))
● where x is in the dimension of the input and output

functions f and g (2D in case of images).

● For discrete images, the domain consists of a finite

number of pixel locations, x = (i, j), and we can

write as:
g(i, j) = h(f(i, j))
1. Multiplication and addition with a constant.
● Two commonly used point processes are
multiplication and addition with a constants:
g(x) = a*f(x) + b
● The parameters a > 0 and b are often called the gain

and bias parameters;

● Sometimes these parameters are said to control

contrast and brightness, respectively.

● The bias and gain parameters can also be spatially

varying,
g(x) = a(x)*f (x) + b(x)
1. Multiplication and addition with a constant.

For a = 1.4; b = 30
2. Dyadic (two-input) operator
● Another commonly used dyadic (two-input) operator
is the linear blend operator.
g(x) = (1 − α)f0(x) + αf1(x).
● By varying α from 0 → 1, this operator can be used

to perform a temporal cross-dissolve between two

images or videos, as seen in slide shows and film
production, or as a component of image morphing
algorithms.
3. Gamma Correction (power law):
● Gamma correction is a nonlinear process that adjusts the
brightness of images to match how humans perceive light.
● Gamma correction applies a power function to each pixel
value in an image.
● The relationship between the input Vignetting
signal brightness Y and
the transmitted signal Y` is given by Y` = Y1/γ. or
g(x) = [f(x)]1/γ
● Gamma values less than 1 make the image darker.
● Gamma values greater than 1 make the image lighter.
● A gamma value of 1 has no effect on the input image.
● where a gamma value of γ ≈ 2.2 is a reasonable fit for most
digital cameras.
Gamma correction: Example

For γ = 2.2
Color transforms: Image composition and
matting
● Image matting is a technique that separates an object from
its background by estimating the transparency of each pixel
in an image.
● Vignetting
It's a key technique in image processing and is used in many
image and video editing applications.
● Formally, matting techniques take as input in image F,
which is assumed to be a convex combination of a
foreground image and a background image B as:
C = (1 − α)B + αF
● where α is the pixel's foreground opacity or matte.
Image matting and composition

Vignetting
Image Histogram:
● Image histogram shows frequency of pixel intensity values.
● x axis shows the gray level intensities

● y axis shows the frequency of intensities.

Vignetting

For 8 bits image, we have 256 levels of gray shades

Image transformation: Thresholding
Original
image

Vignetting
Histogram sliding
Brightness is changed by shifting the histogram to
left or right.

+50
Histogram equalization
●
Histogram equalization is used for enhancing the
contrast of the images.
●
The first two steps are calculating the PDF and
CDF.
●
All pixel values of the image will be equalized.
Histogram equalization
Image with its histogram.
Sample image with 8x8

Small image Image intensity values

sample
Step-1: Find the frequencies of each pixel

Image details

Frequency of pixel values

Step-2 Estimate CDF of each pixel intensity
Step-3 Compute h(v) for each pixel intensity
●
Min = 52
●
Max = 154

cdfmin is the minimum non-zero value of the cumulative distribution

function (in this case 1), M × N gives the image's number of pixels (for
the example above 64, where M is width and N the height) and L is the
number of grey levels used (in most cases, like this one, 256).
Histogram equalization
New min. value = 0, old min. value 52
New max. value = 255, old max. value 154

Original Equalized
Histogram equalization

The same image after

An unequalized image
histogram equalization
Linear filtering
●
A linear filter is a mathematical operation that modifies
an image by changing the signal's frequency spectrum.
●
It's a powerful image enhancement tool that's used to
smooth images, remove noise, and detect edges.
●
Neighborhood operators can be used to filter images to
add soft blur, sharpen details, accentuate edges, or
remove noise.
●
In this section, we look at linear filtering operators,
which involve fixed weighted combinations of pixels in
small neighborhoods.
Understanding Neighborhood in Images
●
In image processing, a "neighborhood" refers to a
group of pixels surrounding a specific pixel.
Correlation filtering
●
The most widely used type of neighborhood operator
is a linear filter, where an output pixel’s value is a
weighted sum of pixel values within a small
neighborhood N (Figure 3.10).

●
The entries in the weight kernel or mask h(k, l) are
often called the filter coefficients.
●
The above correlation operator can be more
compactly notated as g = f⊗h.
Neighborhood filtering

(65x0.1)+(98x0.1)+(123x0.1)+(65x0.1)+(96x0.2)+(115x0.1)+(63x0.1)+(91x0.1)+(107x0.1) = 92
Convolution filtering
●
A common variant of linear filter is convolution
operator.

●
where the sign of the offsets in f has been reversed,
This is called the convolution operator g = f∗h,
●
and h is then called the impulse response function.
Neighborhood operations: Examples

Some neighborhood operations: (a) original image; (b) blurred; (c)

sharpened; (d) smoothed with edge-preserving filter;
Neighborhood operations: Examples

Some neighborhood operations: (e) binary image; (f) dilated;

(g) distance transform; (h) connected components
Padding (Border Effects)
●
We notice that the correlation and convolution operation
produces a result that is smaller than the original image,
which may not be desirable in many applications.
●
This is because the neighborhoods of typical correlation
operations extend beyond the image boundaries near the
edges, and so the filtered images suffer from boundary
effects.
●
To deal with this, a number of different padding or
extension modes have been developed for neighborhood
operations.
Border Effects: Solutions
●
Zero: Set all pixels outside the source image to 0 (a good
choice for alpha-matted cutout images);
●
Constant (border color): Set all pixels outside the source
image to a specified border value;
●
Clamp (replicate or clamp to edge): Repeat edge pixels
indefinitely;
●
(Cyclic) wrap (repeat or tile): Loop “around” the image in a
“toroidal” configuration;
●
Mirror: Reflect pixels across the image edge;
●
Extend: Extend the signal by subtracting the mirrored version
of the signal from the edge pixel value.
Border Effects: Solutions
Separable Filtering
●
The process of performing a convolution requires K2
(multiply-add) operations per pixel, where K is the
size (width or height) of the convolution kernel.
●
This operation can be significantly speed up by first
performing a 1D horizontal convolution followed by
a 1D vertical convolution, which requires a total of
2K operations per pixel.
●
A convolution kernel for which this is possible is
said to be separable.
Separable Filtering
Example Filters
1)Moving average or box filter: Averages the pixel
values in a K × K window.
Box filter: Example
Example Filters
2) Bilinear (Bartlett) filter: A smoother image can
be obtained by separably convolving the image with a
piecewise linear “tent” function.
●
A 3x3 version of bilinear filter is shown below.
Bilinear filter: Example
Example Filters
3) Gaussian filter: Convolving the linear tent
function with itself yields the cubic approximating
spline, which is called the “Gaussian” kernel.
●
A 3x3 version of Gaussian filter is shown below.
Gaussian filter: Example
Example Filters
4) Sobel filter: Linear filtering can also be used as a
pre-processing stage to edge extraction and interest
point detection.
●
A 3x3 version of Sobel filter is shown below.
Sobel filter: Example
Example Filters
5) Corner detector: The simple corner detector
looks for simultaneous horizontal and vertical second
derivatives.
●
A 3x3 version of Corner detector is shown below.
Band-pass and steerable filters
●
More sophisticated kernels can be created by first
smoothing the image with a Gaussian filter.

●
Such filters are known collectively as band-pass
filters, since they filter out both low and high
frequencies.
●
The (undirected) second derivative of a two-
dimensional image,
Band-pass and steerable filters
●
This is known as the Laplacian operator.
●
Blurring an image with a Gaussian and then taking
its Laplacian is equivalent to convolving directly
with the Laplacian of Gaussian (LoG) filter.
Laplacian of Gaussian Filter (LoG)
Second order steerable filter