ADVANCES IN IMAGE

PROCESSING
By,
Prof. Shruthi M. L. J.
Asst. Prof.
Dept. of TCE, CMRIT.
Module 1
The image, its representations and properties:
• Image representations, a few concepts
• Image digitization
• Digital image properties
• Color images.

2
Image representations, a few concepts

• A Signal
• One dimensional
• Two dimensional
• Three dimensional or higher
• A Function can be
• Continuous
• Discrete
• Digital

3
MONOCHROMATIC IMAGES

4
Perspective projection geometry
X = [ x ,y ,z ]T

U = [ u ,v ] T

X
y

v x
u

5
Quality of a digital image
• Grows in proportion to
• Spatial resolution – proximity of image samples in image
plane
• Spectral resolution – bandwidth of the light frequencies
captured by sensor
• Radiometric resolution – number of distinguishable gray
levels
• Time resolution – interval between time samples at which
images are captured.
6
IMAGE DIGITIZATION
• An Image - discrete data structure called matrix.
• An Image by a sensor – continuous function f(x,y)
• Image digitization – function f(x,y) is sampled into a matrix of M rows
and N columns.
• Image quantization – assigns integer value to each continuous sample
i.e. R of f(x,y) is split into K intervals.
• Two aspects of image function sampling
• Sampling period should be determined.
• Geometric arrangement of the sampling points should be set.

7
Sampling
• For most of the applications, quality comparable to ordinary television image is
required.

grid

• Sampling into 512x512 grid is used

8
Quantization
• The transition between continuous values of the image
function(brightness) and its digital equivalent
• Number of quantization levels should be high
• If b bits are used to express brightness values of pixel, then the
number of quantization levels k = 2𝑏
• Problem of images quantized with insufficient quantization levels:
occurrence of false contours
• An efficient computer representation of brightness values – one, two
or eight bits per pixel

9
Quality of image for different quantization
values

10
DIGITAL IMAGE PROPERTIES
• Metric and Topological properties
• Distance : Any function D holding the following three conditions
• 𝑫 𝒑, 𝒒 ≥ 𝟎, 𝒇𝒐𝒓 𝑫 𝒑, 𝒒 = 𝟎 𝒊𝒇𝒇 𝒑 = 𝒒 identity
• 𝑫 𝒑, 𝒒 = 𝑫 𝒒, 𝒑 symmetry
• 𝑫 𝒑, 𝒓 ≤ 𝑫 𝒑, 𝒒 + 𝑫(𝒒, 𝒓) triangular inequality
• Distance metrics:
• Euclidean distance 𝑫𝑬 𝒊, 𝒋 , 𝒉, 𝒌 = (𝒊 − 𝒉)𝟐 +(𝒋 − 𝒌)𝟐
• Advantage: intuitively obvious
• Disadvantages: costly calculation due to square root and its non-integer value

11
𝐷4 or Manhattan or City Block distance
• This is also called 𝐿1 metric
• 𝐷4 𝑖, 𝑗 , ℎ, 𝑘 = 𝑖−ℎ + 𝑗−𝑘

𝐷8 or Chessboard distance
• Equal to the number of minimal king-moves on the chessboard from one
part to another
• 𝐷8 𝑖, 𝑗 , ℎ, 𝑘 = max{ 𝑖 − ℎ + 𝑗 − 𝑘 }

12
Pixel Adjacency
• 4 – neighbours: if 𝐷4 𝑝, 𝑞 = 1

• 8 – neighbours: if 𝐷8 𝑝, 𝑞 = 1

13
Regions and Path
• Path : from pixel P to pixel Q, the path is a sequence of points 𝐴1 , 𝐴2 ,
…….., 𝐴𝑛
Where 𝐴1 = P, 𝐴𝑛 =Q and 𝐴𝑖+1 is a neighbour of 𝐴𝑖
• Region: A set of pixels in which there is a path between any pair of its
pixels
• contiguous: If there is a path between two pixels in the set of pixels in
the image
• The relation “to be contiguous” defines a decomposition of the set
into equivalence classes

14
To be contiguous - decomposed

15
• Assume 𝑅𝑖 - disjoint regions in the image created by “to be
contiguous”
• Let 𝑅 be union of all regions 𝑅𝑖
• 𝑅𝑐 - set complement of 𝑅 is called the background
• Remainder of 𝑅𝑐 is holes
• Simple contiguous: A region with no holes
• Multiple contiguous: A region with holes

16
Paradoxes – due to neighbourhood and
contiguity definitions on square grid
• Two digital segments with 45𝑜 slope

• 4 – connectivity: lines are not contiguous at each of their points

17
• Each closed curve divides the plane into two non-contiguous regions

• Digitization due to 8 – connectivity: A line can be drawn from inner

18
Solutions to paradoxes
• Treat objects using 4 – neighbourhoods and background using 8 –
neighbourhoods or vice-versa
• To use discrete topology based on cellular complexes
• These problems are common in square grids. Hexagonal grids solves
most of these issues

19
Distance function or Chamfering algorithm
Distance transform provides distance of pixels from some image subset

20
• First pass: top left corner of the image and moves horizontally left to
right until bounds of the image is reached and returns to the
beginning of the next row
• Second pass: bottom right corner in the bottom-up right to left
direction using a different local mask

21
22
• Edge: local property of a pixel and its immediate neighbourhood. It is
a vector given by a magnitude and direction
• Crack edge: creates a structure between pixels in a similar manner to
that of cellular complexes

• Border: the set of pixels within the region that have one or more
neighbors outside R

23
• Convex: If any two points within a region are connected by a straight
line segment and the whole line lies within the region
• Convex hull:

24
Histograms

25
Entropy

26
Visual perception of the image
• Contrast: local change in brightness. Defined as ratio between
average brightness of an object and the background.

• Acuity: ability to detect details in an image

27
Perceptual grouping
• A principle used in computer vision to aggregate elements provided
by low-level operations

28