Scribd Logo
Upload

Welcome to Scribd!

  • 49 views

    MCT-453 - MV - 2013 - Lecture #25

    Uploaded by

    MA Khan
    This lecture discusses image analysis techniques. It begins by defining image analysis as automatically extracting useful information from an image. It then discusses common techniques like template matching, statistical pattern recognition, and the Hough transform. Template matching involves finding a pre-defined template within an image and is used for inspection, location, and identification problems. Template matching can be global, using the full template, or local, using smaller templates to detect features. Different similarity measures are used to compare templates to images, such as Euclidean distance, absolute difference, and cross-correlation. Local template matching helps address issues with scale and orientation variations between templates and images.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    MCT-453 - MV - 2013 - Lecture #25

    Uploaded by

    MA Khan
    49 views17 pages
    This lecture discusses image analysis techniques. It begins by defining image analysis as automatically extracting useful information from an image. It then discusses common techniques like template matching, statistical pattern recognition, and the Hough transform. Template matching involves finding a pre-defined template within an image and is used for inspection, location, and identification problems. Template matching can be global, using the full template, or local, using smaller templates to detect features. Different similarity measures are used to compare templates to images, such as Euclidean distance, absolute difference, and cross-correlation. Local template matching helps address issues with scale and orientation variations between templates and images.

    Original Description:

    Machine vision lecture -25

    Original Title

    MCT-453_MV_2013_Lecture #25

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    This lecture discusses image analysis techniques. It begins by defining image analysis as automatically extracting useful information from an image. It then discusses common techniques like template matching, statistical pattern recognition, and the Hough transform. Template matching involves finding a pre-defined template within an image and is used for inspection, location, and identification problems. Template matching can be global, using the full template, or local, using smaller templates to detect features. Different similarity measures are used to compare templates to images, such as Euclidean distance, absolute difference, and cross-correlation. Local template matching helps address issues with scale and orientation variations between templates and images.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    49 views17 pages

    MCT-453 - MV - 2013 - Lecture #25

    Uploaded by

    MA Khan
    This lecture discusses image analysis techniques. It begins by defining image analysis as automatically extracting useful information from an image. It then discusses common techniques like template matching, statistical pattern recognition, and the Hough transform. Template matching involves finding a pre-defined template within an image and is used for inspection, location, and identification problems. Template matching can be global, using the full template, or local, using smaller templates to detect features. Different similarity measures are used to compare templates to images, such as Euclidean distance, absolute difference, and cross-correlation. Local template matching helps address issues with scale and orientation variations between templates and images.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    of 17

    Lecture # 25: Image Analysis

    Muhammad Rzi Abbas Department of Mechatronics and Control Engineering


    [email protected]
    Lecturer, Mechatronics Dept.
    University of Engineering and Technology, Lahore
    • The term Image Analysis embodies the idea of automatically
    extracting useful information from an image of a scene.
    • Techniques vary across a broad spectrum, depending on the
    complexity of the image and, indeed, on the complexity of the
    information to be extracted from it.
    • Commonly used image analysis techniques include Template
    Matching, Statistical Pattern Recognition, and the Hough Transform.
    • More useful classification can be done based on the function.
    • Inspection: More concerned with visual appearance
    • Location: Localization of the object with respect to the image frame of
    reference
    • Identification: Classification of unknown objects
    • Generally speaking,
    • Inspection applications utilize the template matching paradigm
    • Location problems utilize the template matching paradigm and Hough
    Transform
    • Identification problem normally utilize all three techniques, depending on the
    complexity of the image and the objects
    • Many of the applications of computer vision simply need to know
    whether an image contains some previously defined object or, in
    particular, whether a pre-defined sub-image is contained within a test
    image.
    • This sub-image is called a template and should be an ideal
    representation of the pattern or object which is being sought in the
    image.
    • A measure of matching between the template and every location of
    the image is calculated and decision is made based on its value.
    • There are two forms of template matching
    • Global Template Matching
    • Template used is a global representation of the object
    • Local Template Matching
    • Several templates are used which represent local features of the object,
    e.g. corners, edges, characteristic marks etc. present in the object
    Measure of Similarity
    • Several similarity measures are possible, some based on the
    summation of differences between the image template, others based
    on the cross-correlation techniques.
    Measure of Similarity
    • Metric Based Measure
    • Euclidean Distance Measure is defined as

    • g(i,j) = Image
    • t(i,j) = Template
    Measure of Similarity
    • Metric Based Measure
    • Absolute Difference Measure

    • Squared Difference Measure


    Measure of Similarity
    • Cross-Correlation Measure
    • Suppose we remove the constant portions from the Squared Difference
    Measure we will be left with what is known as Cross-Correlation Measure
    Measure of Similarity
    • Cross-Correlation Measure
    • Normalized Cross-Correlation is a more useful measure
    Local Template Matching
    • One problem of template matching is that each template represents
    the object or part of it as we expect it to appear in the image.
    • No recognition is given to the variation in scale or in orientation.
    • If the expected orientation can vary, then we will require a separate
    template for each orientation and each one must be matching with
    the image.
    • Thus template matching may become computationally expensive,
    especially if the templates are large.
    Local Template Matching
    • One popular way of alleviating this computational overhead is to use
    much smaller local templates to detect salient features in the image
    which characterize the object we are looking for.
    • Machine Vision by David Vernon, Published in 1991
    • Chapter 6 (Section 6.1 and 6.2)
    • www.cs.nccu.edu/~xhuang/cisg5740-f10/18-recognition.pdf

    You might also like