Transcript Machine Vision

Digital Image Processing
&
Machine Vision
By: D.K.SONI
Fundamental Steps in DIP
Components of Image Processing
Systems
Fundamental Steps
• Low level Process
• Medium level process
Low level Processes
• These are the methods in which both input
and output are images
• Ex. Contrast enhancement
Medium Level Processes
• These are the method in which inputs may
be images but outputs are attribute
extracted from image.
• Ex. Partitioning an image in to regions &
recognition of individual object
License Plate reading
Image Acquisition
• How to obtain a good image?
• The Properly taken image can reduce lots
of steps required to obtain the final image.
• Brightness & contrast play an important
role in image acquisition.
Methods of Image Acquisition
Digital Image?
Sampling & Quantization
Gray-level resolution
• Number of bits used to represent intensity
of each pixel.
• Ex. 8 bits , 16 bits
• An image with 256 possible greylevel
values is called an 8-bit image.
• L =2k
Image Enhancement: 3 & 4
• Objective of enhancement is to process an
image so that the result is more suitable
than the original image for a specific
application.
• Means a technique which is suitable for xray may not be suitable for gamma ray.
Contrast Enhancement
Enhancement
• Very subjective area of image processing.
• Process are done in spatial & frequency
domain.
• Spatial means modifying directly the pixel
values.
• Frequency domain means modifying the
Fourier transform of an image.
Spatial Enhancement Processes
• How to obtain negative of an
image?
• How to adjust the gamma
value in image for better
display?
Gamma correction
• Adjusting contrast
• Contribution of each bit in the image
Histogram processing
• Gray level value on the Horizontal axis &
no. of pixels on the vertical axis.
Logical & Arithmetical Operations
Image Addition & subtraction
Filtering in Spatial Domain
• Filter is nothing but one type of matrix.
• Image is also matrix.
• So filtering process means performing
mathematical calculations between filter &
image to remove unwanted data (Noise)
from an image or to sharpen the image.
Removing noise using filter
Sharpening
Image Restoration, 5
• These are the processes used to improve
the appearance of an image.
• Objective in nature
• Based on some predefined rule.
Compression
• Techniques: reduce the size required to
save an image or the bandwidth required
to transmit it.
• Ex. jpeg
Morphological Image Processing
• Extracting useful image components that
are useful in the representation and
description of shape.
Knowledge Base
• Knowledge about the problem domain is
stored in an Image processing system in
the form of a knowledge base.
• It controls the interaction between different
modules.
Components of an IPS
DIGITAL
IMAGE
FUNDAMENTALS
What Makes a good image?
• Cameras (resolution, focus),
• Distance from object (field of view),
• Illumination (intensity of light, direction,
expose time, light source),
• Background characteristics
• etc
Human Visual Perception
• Why is it important?
• Because many techniques in DIP are
highly subjective.
• Human analysis plays a central role
• Structure of the Human eye
• Human vision & Computer Vision.
• Effects of brightness & contrast on human
vision.
Structure of the Human Eye
• We can see any object because the light is
reflected from that object.
• The reflected light is focused on the retina.
• Retina contains light receptors over its
surface. These receptors are like sensors.
• There are two classes of receptors: cones
and rods
Cones
• Primarily responsible for colour perception.
• Photopic (high light) vision: requires higher intensity of
light.
• 6-7millions, located primarily in the central part of the
retina called fovea
• muscles controlling the eye rotate the eyeball until the
image of an object of interest falls on the fovea.
• Each is connected to its own nerve, like each
photoelectric cell is connected to digitizing circuit through
separate wire.
• human can resolve the fine detail.
Rods
• Scotopic (low light) vision.
• 75-150millions, distributed over the retina
surface.
• Several rods are connected to a single nerve
end so reduce the amount of details transferred.
• Serve to give a general, overall picture of the
field of view.
• 10 times more sensitive to the light than cones.
• Mainly sensitive to low illumination (low intensity
of light).
Human & camera vision
• Human Lens is flexible
• When we see object at far distance , lens
becomes flat, to capture maximum light
rays reflected from the object. Same
happens with zoom in feature of camera.
• When we see object close to us, lens
becomes thicker , to capture only rays
reflected from the object. Its like zoom out
in camera.
BRIGHTNESS ADAPTATION
• When you enter a dark theater on a bright
day, it takes an appreciable interval of time
before you can see well enough to find an
empty seat. Why is it so?
Brightness Adaptation
Brightness adaptation
• Human vision system can adapt very wide
range of brightness ranging from scotopic
threshold (very dark environment) to glare
limit (very bright environment).
• But the human vision cannot operate over
such a a range simultaneously.
• Total range of distinct intensity levels it
can discriminate (see separately) is
smaller than total adaptation range.
• Brightness is a function of intensity &
simultaneous contrast.
• It takes more time for eyes to adjust when
we move from brighter to darker
environment.
• But it takes less time when we move from
dark to light environment.
Brightness is a function of intensity
Brightness is a function of contrast
Image Acquisition
• Using Single Sensor
• Line sensor
• Array sensor
Image Sensor
• Simple Image sensor is photodiode.
• Constructed from CCD or CMOS
• When light strikes, it gives some voltage
as output based on intensity.
• Filter is used with sensor for selectivity.
How to Obtain 2D image with
single sensor?
• There must be relative displacement in
both the x- and y direction between the
sensor & area to be imaged.
• Use rotation of a drum and lead screw
technique.
• Cheaper but slower method for image
acquisition.
Image Acquisition using Line
sensor
• The strip provides imaging in the one
direction , while motion perpendicular to
strip provided imaging in the other
direction.
• Used in Flat-Bed scanners
• Also used in tomography.
Image Acquisition using Sensor
array
• Most widely used
• Entire image can be scanned without relative
displacement.
• Light rays reflected from the object are focused
on the image plane (sensor array).
• Digitizing circuitry converts output voltages of
each sensor in to corresponding number. As a
result digital output is obtained.
• The process of digitizing of image involves 2
main steps. i.e. sampling & quantization
Digitizer
Image formation model
Image Sampling and
Quantization
Digital Images as Matrices
A simple Image formation model
• Monochromatic Images: Images which ranges
from black to gray to white.
• Monochromatic images have different shades of
gray.
• Digital image is a 2D function f(x,y)
• Where f is the intensity or gray level
• (x,y) are spatial coordinates
• For a digital image f(x,y) must be finite
• Since f(x,y) represent the intensity of light
reflected from the object. If it is zero
means there is no radiation of the energy
(light) from the object.
• Means there is no image.
• So 0 < f(x,y) < ∞
•
Value of F(x,y) depends on two factors.
1. How much amount of light from the
Illuminating source strikes on the object.
2. Amount of Illumination reflected by the
object.
These are called illumination i(x,y) & reflection
r(x,y) components respectively.
•
•
•
•
F(x,y) = i(x,y) * r(x,y)
Where
0 < i(x,y) < ∞ and 0 < r(x,y) <1
r= 0.01 for black velvet, 0.9 for silver
plated metal & 0.93 for snow.
l=f(x0,y0)= gray level
Lmin≤ l ≤ Lmax
Image Sampling
• Output of most sensors is a continuous
voltage waveform from sensed data.
• Conversion of these data in the digital
form involves 2 main processes
• Image sampling & Quantization
Sampling & Quantization
• Quality of a digital image depends on the
number of samples & discrete gray levels
used in sampling & quantization.
• That’s why 10 MP camera gives you better
quality of picture than 2 MP.
Digital Image as Matrix
APPLICATIONS OF DIP
Examples of Fields that Use DIP
Gamma-Ray Imaging
X-ray Imaging
THE SUN
THE Venus
Imaging in Visible & Infrared Band
Taxol (Anti Cancer
Agent – 250 x)
Brazil,
Vegetation
Hurricane,
Andrew
America
Africa Europe
Russia India
Australia
Ultrasound Images
MRI
Mountains in southeast Tibet
Automated License Plate Reading
Paper Currency
Thumb Print
Bubbles in Plastic Product
A circuit Board Controller
Packaged Pills
Bottles
THANK YOU FOR
YOUR
ATTENTION