Software systems for Computer Vision and Image Processing

companyscourgeAI and Robotics

Oct 19, 2013 (3 years and 10 months ago)

84 views

Software systems

for Computer Vision and Image Processing

Sungsoo

Ha


Prof.
Murali

Subbarao

(Stony Brook University)

Prof. A.N.
Rajagopalan

(Indian Institute of Technology, Madras, India)

Arnav

V.
Bhavsar

(PhD student, Indian Institute of Technology, Madras, India)

Contents


Machine (or Computer) Vision System


Image Processing


Early Vision Process


Edge detection (the Canny Edge detection)


Intermediate Vision Process


Hough Transform


High Vision Process


Computed Tomography (Filtered
backprojection

algorithm)

Machine Vision System


Image Processing


Early Vision Process


Intermediate Vision Process


High Vision Process

Image Processing

Enhancing Image Contrast

Reducing Noise

Smoothing

Edge detection

Arbitrary Filter

Linear filter: Convolution

The Canny Edge Detector


Edge in an image


Significant local changes in an image


Important features for analyzing image


Canny Edge Detector


The optimal edge detector


Low error rate


Well localized edge points


Only one response to a single edge

Cont’d (Canny Edge detector)


Algorithm

1.
Filtering:

Smooth the image

2.
Enhancement
: Compute the gradient magnitude and orientation

3.
Detection
: Apply non
-
maxima suppression

4.
Localization
: Use double
thresholding

Hough Transform (HT)

for line detection


What is difference with edge detector?


Application: geometric pattern matching

1.
An image of a single object

2.
Decomposed into lines, curves, or other shapes

3.
Matched with those in the desired object

Computed Tomography (CT)

CT : the cross
-
sectional
imaging of an object
from its projection data

Parallel beam
projection

Filtered back
-
projection
algorithm

3D Volume Rendering

&

Graphic User Interface (GUI)


System & Software Requirement


OpenGL extension version (2.0)


Cg toolkit (the latest version)


GeForce

8000 series (at least)




Conclusion


Summarize


Realize very basic and simple applications


Help to understand overall of machine vision
system


Future work


Improving Hough Transform to detect arbitrary
curves


Medical Image Processing


Two
-
level
volume rendering


SPECT and PET statistical image reconstruction