Binary Image Processing

By: Amin Khajeh Djahromi

Department of Electrical Engineering University of Texas at Arlington

Roadmap
Binary Image Processing Thresholding Geometric Properties Run-Length Encoding Binary Algorithms Morphological Operators Applications

What is Binary Image?

Binary images are images whose pixels have only two possible intensity values.

Binary Image

Gray Level Image

Why do we use binary images?

Smaller memory requirements Faster execution time Many techniques developed for these systems are also applicable to vision systems which use gray scale images Less expensive

Disadvantages of using binary images

Limited application Can not be extended to 3D Losing internal details of objects (i.e. in inspection tasks) difficult to control the contrast between the background and the objects (i.e. in material handling and assembly tasks)

Some aspects of Binary Vision Systems

Formation of binary images : Binary is obtained by thresholding the gray scale image. Many cameras are designed to perform thresholding in hardware. Geometric Properties: Size, Position, Orientation and Projection. Topological Properties Object recognition in binary images

Segmentation
Segmentation: The partitioning of an image into regions is called segmentation

Thresholding
Thresholding: Is a method to convert a Gray Scale Image into a Binary Image, so that objects of interest are separated from the background.

Example of Thresholding

Original

Binary (Threshold =140)

Importance of object-background separation

For thresholding to be effective in objectbackground separation, it is necessary that the objects and background have sufficient contrast and that we know the intensity levels of either the objects or the background. In a fixed thresholding scheme these intensity characteristics determine the value of threshold.

Some Types of Thresholding

Fixed Thresholding Histogram-derived Thresholds Isodata algorithm Background Symmetry algorithm Triangular algorithm

Geometric properties
Size Position Orientation Projection

Size
Size: summation of all
pixel values or the number of pixels of the object.

0 0 0 0

0 1 1 0

0 1 1 0

0 0 0 0

Size = 4

Position: The center of the area of the object. In order to

calculate the position of the object we need to calculate the firstorder moments.

Orientation: The way an object is situated in the image. For

some shapes (such as circle), orientation is not unique and to define unique orientation an object must be elongated.

Projection:
1. 2. 3.

Horizontal projection Vertical projection Diagonal projection

Run-length Encoding (RLE)

Compact representation of a binary image Used for image transmission Some properties, such as the area, projection of the objects, may be directly computed from their run-length codes. Greater compression achieved for longer and more frequent runs Widely used for binary images.

General example for RLE

AAAAABBBBB require 10 bytes to store Run Counts : 5 5 Run Value : A B RLE code, 5A5B RLE require only 4 bytes Compression ratio : 10/4=2.5

Another example of RLE

Input : DDDDBBBBGVVVVVV RLE codes: 4D4B1G6V RLE Packet : 8 bytes Compression ratio : 15/8

Two approaches for RLE

1.

Using start position and length of 1s for each row. Using length of runs, starting with the length of the 1 run for each row.

2.

Example for approach 1

1 1 1 1 0 0 0 1 1 1 1 1 1 1 0 0 0 0 0 0 1 1 1 1 1 0 0 0 0 0 0 0 1 1
Start position and length of 1 runs: (1,4) (8,7) (4,5) (16,2)

Example for approach 1

11110000011111000 00001111110000011

Length of 1 and 0 runs (starting with the length of 1 runs) :

4,5,5,3 0,4,6,5,2

Binary Algorithms
Before discussing Binary Algorithms, we first need to define some related terms. Neighbors Path Foreground Background Connectivity Connected Components Boundary Interior Surrounds

Neighborhood
A pixel has a common boundary with four pixels and shares a corner with four additional pixels. 4-neighbors: Two pixels are 4-neighbors if they share a common boundary. 8-neighbors: Two pixels are 8-neighbors if they share at least one corner.

Examples of 4-neighbors and 8-neighbors

(a) 4-neighbors, (b) 8-neighbors

Path
A path from the pixel at [i ] is a sequence of pixel indices [i0,j0],[i1,j1],[in,jn] such that the pixel at [ik,jk] is neighbor of the pixel at [ik+1,jk+1] for all k with 0 < k < n-1 . According to the above definition we will have to different paths: 1. 4-path : using 4-connection 2. 8-path : using 8-connection
0,j0

Examples of 4-path and 8-path

(a) 4-path

(b) 8-path

Foreground, Background and Holes

Foreground: The set of all 1 pixels in an image is called the foreground and is denoted by S Background: The set of all connected components of complement of S that have points on the border of an image is called the background. All the other components of complement of S are called holes.

Example of foreground and background

Consider the below picture:

1 1 1
If we consider 4-connectedness for both foreground and background, there are four Objects that are 1 pixel in size and there is no hole. But, if we use 8-connectedness, then there is one object and no hole. So to avoid this awkward situations, if we use 8-connectedness for foreground, then 4-connecteness should be used for background.

Connectivity
A pixel p S is said to be connected to q S if there is a path from p to q consisting entirely of pixels of S. For any three pixels p,q and r in S, we have: 1. Pixel p is connected to p (reflexivity). 2. If p is connected to q, then q is connected to p (commutativity). 3. If p is connected to q and q is connected to r, then p is connected to r (transitivity).

Example of connectivity

Two connected components based on 4-connectivity

Boundary, Interior and surrounds

Boundary: The boundary of S is the set of pixels of S that have 4-neighbors in Sbar. The boundary is usually denoted by S` . Interior: The interior is the set of pixels of S that are not in its boundary. The interior of S is (S-S`). Surrounds: T surrounds region S, if any 4path from any points of S to the border of picture must intersect T.

Example of Boundary, Interior and surrounds

Original image

Black: boundary pixels Gray: interior pixels White: surrounds pixels

Components Labeling
To find connected components in an image. The points in a connected components form a region that represent an object. Require the characteristics of the components (size, position, orientation and )

Example of Component Labeling

Original image

Its connected components

Two algorithm for Component Labeling

1- Recursive algorithm Rarely used on general-purpose computers Commonly used on parallel machines 2- Sequential algorithm Usually require two passes over the image Works with only two rows of an image at a time so there is no need to bring the whole image into memory

Recursive algorithm
Scan the image to find an unlabeled 1 pixel and assign it a new label L. 2. Recursively assign a label L to all its 1 neighbors. 3. Stop if there are no more unlabeled 1 pixel. 4. Go to step 1.
1.

Sequential algorithm
1- Scan the image left to right, top to bottom 2- If this pixel is one
(a) if only one of its upper and left neighbors has a label, then copy the label (b) If both have the same label, then copy the label. (c) If both have different labels, then copy the uppers label and enter the labels in the equivalence table as equivalent labels (d) Otherwise assign a new label to this pixel and enter this label in the equivalence table.

3- If there are more pixels to consider, then go to step 2. 4- Find the lowest label for each equivalent set in the equivalence table. 5- Scan the picture. Replace each label by the lowest label in its equivalent table.

Example of Component Labeling

Size filtering
If we know that objects of interest are of size greater than T0 , then we can use size filter to remove noise after component labeling. In this case all components below T0 in size are removed by changing the corresponding pixels to 0.

Example of size filtering

Original image ( noisy letter i )

size filtered image T=3

Euler number
E (Euler No.) used as a feature of an object. E= C H Where C = number of connected components (objects) and H = number of Holes. E is invariant to translation, rotation and scaling. (fixed feature of an object)

Example of Euler number

For this image: C (number of objects)=? H (number of holes) =? E=C-H = ? C=1 !!! H=3 !!! E= -2

Distance measure
Euclidean 2. City-block 3. Chessboard The city-block and chessboard measure are preferred over Euclidean due to their simplicity of calculation. Euclidean distance is closest to the continue case, but it is computationally the most expensive.
1.

Distance transform
Representing an image in terms of distance. It is useful in character recognition. It can calculated by using a parallel iterative algorithm

Distance Measure -Euclidean

The Euclidean distance is the straight line between two pixels.

Distance Measure -Cityblock

Measure the path between the pixels based on a 4-connected neighborhood. Pixels whose edges touch are 1 unit apart. Pixels diagnally touching are 2 units apart.

Distance Measure -Chessboard

Measure the path between the pixels based on an 8-connected neighborhood. Pixels whose edges or corners touch are 1 unit apart.

Medial Axis/ Skeletonization

Used to reduce foreground regions in a binary image to a skeletal remnant that largely preserves the extent and connectivity of the original region while throwing away most of the original foreground pixels. Methods of skeletonization:
Thinning Calculate distance transform of the image -the skeleton then lies along the singularities (i.e. creases or curvature discontinuities) in the distance transform.

Example

Skeletonization

Thinning
Thinning is an image processing operation in which binary-valued image regions are reduced to lines that approximate their center lines. Also called as skeleton or core Reduce image components to their essential information

Thinning requirements
1. 2. 3. 4. 5.

Connected image regions must thin to connected line structures. The thinned result should be minimally 8connected. Approximate endline location should be maintained. The thinning results should approximate the medial lines. Short branches caused by thinning should be minimized.

Example of Thinning

Threshold = 60

Expanding and shrinking

Expanding : Change a pixel from 0 to 1 if any neighbors of the pixel are 1. Shrinking : Change a pixel from 1 to 0 if any neighbors of the pixel are 0. According to the above definition, Shrinking may be consider as expanding the background.

Properties of Shrinking and Expanding

Expanding followed by Shrinking can be used for filling undesirable holes. Shrinking followed by Expanding can be used for removing isolated noise pixels. Expanding and Shrinking can be used to determine isolated components and clusters.

Example of Shrinking and Expanding

An example of expanding and shrinking operations on the letter s Left: the original image Middle: Expanded image Right: Shrunken

References (books)
Machine Vision (Ramesh Jain, Rangachar Kasturi, Brian G.
Schunck)

Automated Visual Inspection (B. G. Batchelor, D. A.

Hill, and D. C. Hodgson )

Intelligent Image Processing in Prolog (B. G.

Batchelor )

References (Web)
http://bruce.cs.cf.ac.uk/bruce/Machine_vision_tutorial/Bin ary_image_processing.html University of North Dakota, Department of Computer Science, computer vision website http://www.machinevisiononline.org/public/articles/archiv edetails.cfm?id=1396 Computer Engineering Department, Ko University (course website)