Scribd
Upload
65 views

Texture Segmentation Using Uniform Local Binary Pattern

This document discusses texture segmentation using the Uniform Local Binary Pattern (ULBP) approach. It begins by defining image segmentation and texture segmentation. It then describes the ULBP approach, which examines intensity relationships between neighboring pixels rather than absolute intensity values. An algorithm is provided that calculates ULBP values for image windows and segments the image by comparing ULBP values to sample textures. Experimental results on various images show classification accuracy generally decreases as window size decreases but remains above 75% even for small windows. In conclusion, ULBP gives good results using only 9 patterns and is invariant to grayscale and rotation changes.

Uploaded by

G.m. Raja
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
65 views

Texture Segmentation Using Uniform Local Binary Pattern

This document discusses texture segmentation using the Uniform Local Binary Pattern (ULBP) approach. It begins by defining image segmentation and texture segmentation. It then describes the ULBP approach, which examines intensity relationships between neighboring pixels rather than absolute intensity values. An algorithm is provided that calculates ULBP values for image windows and segments the image by comparing ULBP values to sample textures. Experimental results on various images show classification accuracy generally decreases as window size decreases but remains above 75% even for small windows. In conclusion, ULBP gives good results using only 9 patterns and is invariant to grayscale and rotation changes.

Uploaded by

G.m. Raja
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
You are on page 1/ 28

Texture segmentation using

Uniform Local Binary Pattern


Presented by
G.MADASAMY RAJA

IMAGE SEGMENTATION

Process of splitting an entire image into a set of


regions

Segmentation is useful for


Image Analysis for which it is considered as initial step
Image compression since homogeneous regions can be
encoded more compactly

IMAGE SEGMENTATION

TEXTURE SEGMENTATION

Partitioning of an image into regions, each


of which contains a single texture distinct
from its neighbors.

TEXTURE

We can define texture as repeating patterns


of some pixel values in image intensity in a
periodic manner.

Texture Segmentation- Procedure


Input Image

Transfer into texture models

Image Segmentation

Display

Stop

ULBP

UNIFORM LOCAL BINARY


PATTERN APPROACH

A Texture image is decomposed into many


patterns

The key concept of ULBP method is the


computation of the intensity relations
between the pixels in a small neighborhood
and not their original intensity values.

UNIFORM LOCAL BINARY


PATTERN APPROACH

Local Binary pattern Approach:

The idea:
Measure texture by examining the 8-neighbors of every pixel p
to see if their intensity is greater than that of p.

The process:
Construct an 8-digit binary number b1b2b3b4b5b6b7b8 such that
bi = 1 if the intensity of the ith neighbor is > that of p and 0
otherwise.

UNIFORM LOCAL BINARY


PATTERN APPROACH
Uniform patterns are identified
Uniformity refers to the patterns that have
at most 2 spatial transitions or bitwise 0/1
changes in the pattern
For example, patterns 00000000 2 and
111111112 have U value of 0

UNIFORM LOCAL BINARY


PATTERN APPROACH
Uniform Patterns

1. 00000000 - 0
2. 11111111 - 0
3. 01000000 - 2
4. 01100000 - 2
5. 01110000 - 2
6. 01111000 - 2
7. 01111100 - 2
8. 01111110 - 2
9. 01111111 - 2

SEGMENTATION THROUGH
UNIFORM LOCAL BINARY PATTERN APPROACH
Algorithm:

Select sample sub image of 30*30 pixels


from each texture image (one sample per
texture)
Calculate ULBP for each sample sub
image, by moving the 3*3 matrix, across
the sample with overlap
Scan the input image by a window of
30*30 pixels and calculate the ULBP for
each window.
Calculate the absolute difference between
the ULBP of each window and one of the
each sample
The central pixel of the window
considered will be assigned to any one of
the sample class k such that D(K) is
minimum

SEGMENTATION RESULTS-ULBP
Experiment #1
Image segmentation by using Uniform
Local Binary Pattern method, when the
window size is 25*25(four-texture image)

(a) Original image

(b) Segmented image

SEGMENTATION RESULTS-ULBP
Experiment #2
Image segmentation by using Uniform
Local Binary Pattern method, when the
window size is 20*20(four-texture image)

(a) Original image

(b) Segmented image

SEGMENTATION RESULTS-ULBP
Experiment #3
Image segmentation by using Uniform
Local Binary Pattern method, when the
window size is 15*15(four-texture image)

(a) Original image

(b) Segmented image

SEGMENTATION RESULTS-ULBP
Experiment #4
Image segmentation by using Uniform
Local Binary Pattern method, when the
window size is 10*10(four-texture image)

(a) Original image

(b) Segmented image

SEGMENTATION RESULTS-ULBP
Experiment #5
Image segmentation by using Uniform
Local Binary Pattern method, when the
window size is 25*25(two-texture image)

(a) Original image

(b) Segmented image

SEGMENTATION RESULTS-ULBP
Experiment #6
Image segmentation by using Uniform
Local Binary Pattern method, when the
window size is 20*20(two-texture image)

(a) Original image

(b) Segmented image

SEGMENTATION RESULTS-ULBP
Experiment #7
Image segmentation by using Uniform
Local Binary Pattern method, when the
window size is 15*15(two-texture image)

(a) Original image

(b) Segmented image

SEGMENTATION RESULTS-ULBP
Experiment #8
Image segmentation by using Uniform
Local Binary Pattern method, when the
window size is 10*10(two-texture image)

(a) Original image

(b) Segmented image

RESULT ANALYSIS-ULBP

This table shows


the classification
Accuracy for the
varying window
size with fourtexture image.

s.no

Window size

Classification
Accuracy

25*25

80%

20*20

78.3%

15*15

76.67%

10*10

68.4%

CLASSIFICATION ACCURACY
VS
WINDOW SIZE ULBP (four-texture)

RESULT ANALYSIS-ULBP

This table shows


the classification
Accuracy for the
varying window
size with twotexture image.

s.no

Window size

Classification
Accuracy

25*25

98%

20*20

97%

15*15

97%

10*10

97%

CLASSIFICATION ACCURACY
VS
WINDOW SIZE ULBP (two-texture)

CONCLUSION-ULBP
Gives better results even when only 9
uniform patterns are considered
Average Classification rate is 86%
(inclusion of boundary values)
Classification rate goes higher if boundary
regions are removed
Invariant against grayscale and rotation

CONCLUSION-ULBP (Grayscale
Invariant)

CONCLUSION-ULBP (Rotation
Invariant)

Applications

Satellite Images Analyzing

Medical Imaging (Endoscope images,


Cancer tracing)

Quality of the Product is checked

You might also like