16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi...

download 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

of 37

Transcript of 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi...

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    1/37

    nuMAPA Content Based Image Retrieval Project

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    2/37

    BSCS Final Year EveningGroup Members

    Mohammad Umer Sheikh EP046125

    Syed Arbab Ahmed EP046142

    Pervaiz Ahmed EP04A6136

    Noman Iqbal EP046133 Mustafa Turab Ali EP04A6132

    Project Supervisor Dr.Aqil Burny

    Badar Sami

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    3/37

    Syed Arbab Ahmed

    EP046142

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    4/37

    Definition

    Content-based image retrieval (CBIR), alsoknown as query by image content (QBIC)and content-based visual information

    retrieval (CBVIR) is the application ofcomputer vision to the image retrievalproblem, that is, the problem of searchingfor digital images in large databases.

    http://e/wiki/Computer_visionhttp://e/wiki/Image_retrievalhttp://e/wiki/Digital_imagehttp://e/wiki/Databasehttp://e/wiki/Databasehttp://e/wiki/Digital_imagehttp://e/wiki/Image_retrievalhttp://e/wiki/Computer_vision
  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    5/37

    Scope of the project

    Content-based image retrieval potentiallyprovides new opportunities to extend andenhance the constraints and limitations imposed

    by the traditional information retrieval paradigmon image collections.

    The number of CBIR systems is extremelyencouraging.

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    6/37

    CBIR Systems

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    7/37

    Potential uses for CBIR include

    Photograph archives

    Retail catalogs

    Medical diagnosis

    Crime prevention The military

    Art collections

    Intellectual property

    Architectural and engineering design

    Geographical information and remote sensingsystems

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    8/37

    Difference b/w human with Computer

    The basic reason why image retrieval is moredifficult than text retrieval is that the digitalrepresentation for most images is as acollection of pixels.

    The only information which is explicit in such arepresentation is the color values at each pixel

    point.

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    9/37

    CBIR software systems and techniques

    Query by example

    Semantic retrieval

    Other query methods

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    10/37

    Pervaiz Ahmed

    EP04A6136

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    11/37

    Our CBIR System Design

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    12/37

    Problem Statement

    The problem involves entering an image as aquery into a software application that is

    designed to employ CBIR techniques inextracting visual properties, and matchingthem. This is done to retrieve images in thedatabase that are visually similar to the

    query image

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    13/37

    Requirement Analysis

    At the very first step we require an algorithmwhich extract features from images.

    SIFT algorithm for features extraction

    NNS for matching

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    14/37

    SIFT Algorithm(Scale-Invariant Feature Transform )

    SIFT is an image processing algorithm whichcan be used to detect distinct features in an

    image. Once features have been detected for two

    different images, one can use these features toanswer questions like are the two images

    taken of the same object?

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    15/37

    Out put of SIFT

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    16/37

    Noman Iqbal

    EP046133

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    17/37

    Algorithm working phases

    Four phases of SIFT

    1 Scale-space Extrema Detection

    2 Key point localization3 Orientation Assignment

    4 Key point descriptor

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    18/37

    Phase 1: Scale-space Extrema Detection

    The first phase of the computation seeks toidentify potential interest points. It searchesover all scales and image locations. The

    computation is accomplished by using adifference-of-Gaussian (DoG) function. Theresulting interest points

    are invariant to scale and rotation, meaning

    that they are persistent across image scalesand rotation.

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    19/37

    Phase 2: Key point localization

    For all interest points found in phase 1, adetailed model is created to determinelocation and scale.

    Key points are selected based on theirstability. A stable key point is thus a key pointresistant to image distortion

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    20/37

    Phase 3: Orientation Assignment

    For each of the key points identified in phase2, SIFT computes the direction of gradientsaround.

    One or more orientations are assigned to

    each key point based on local image gradientdirections.

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    21/37

    Phase 4: Key point descriptor

    The local image gradients are measured in theregion around each key point.

    These are transformed into a representation thatallows for significant levels of local shape

    distortion and change in illumination.

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    22/37

    Mustufa Turab Ali

    EP04A6132

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    23/37

    NNS Algorithm(nearest neighbor search)

    For matching we use a NNS.

    An algorithm that is able to detect similaritiesbetween key points.

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    24/37

    Output of NNS

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    25/37

    KD-tree

    KD-tree is the most importantmultidimensional structure decomposes amultidimensional space into hyper

    rectangles.A binary tree with both a dimension numberand splitting value at each node Each nodecorresponds to a hyper rectangle Fields of

    KD-tree node.

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    26/37

    KD-Tree

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    27/37

    Image matching

    A match where the whole of one imagematches the whole of another image.

    Part of one image matching the whole ofanother image.

    Part of one image matching part ofanother image.

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    28/37

    Image Test 1

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    29/37

    Image Test 2

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    30/37

    Image Test 3

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    31/37

    Muhammad Umer Sheikh

    EP046125

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    32/37

    Key point generation

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    33/37

    Key point matching

    Select a node from the set of all nodes not yetselected.

    Mark the node as selected.

    Locate the two nearest neighbors of the selectednode.If the distance between the two neighbors are lessthan or equal to a given distance, we have a match.

    Mark the key points as match.

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    34/37

    Key points matching

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    35/37

    Quality of Match

    KS the numbers of Key points in sourceimage

    KC the numbers of Key points in compare

    imageKM the numbers of Key points in matchimage

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    36/37

    Summary and Conclusion

    SIFT does what it is designed to do, and itdoes it well. The most obvious drawback withSIFT is the time it takes to compare two

    images. The running time of an NNS search isso large that it effectively renders SIFTuseless for a System like M2S. However, withmodifications like quality of match and the

    utilization of other metadata, SIFT could be anextremely robust resource for object detectionand image matching.

  • 8/2/2019 16580815 Final Presentation of CBIR Through SIFT Algorithm of Our Final Project of BSCS From Karachi University

    37/37

    Thank you

    Questions?