Slide 1Slide 2Slide 3Slide 4Slide 5Slide 6Slide 7Slide 8Slide 9Slide 10Slide 11Slide 12Slide 13Slide 14Slide 15Slide 16Slide 17Detecting Edges in ImagesDetecting Edges in ImagesBy Sean SzumlanskiBy Sean SzumlanskiAgenda•Reasons for detecting edges•Introduction to computer images•Defining the edge•Methodologies for edge detectionWhy edge detection?•Segmenting into areas and objects of interest•Psychological processing of images•Tracking moving objects.Introduction to images•Images as 2D arrays of numbers•Color images: RGB values•Black & White images: intensity values•Intensity = (R + G + B) / 3What is an edge?•Heightened rate of change of intensity •High magnitude of gradientEdge detection process•Get rid of noise: Gaussian Filter•Greater sigma value means smoother image•Convolve image with mask•Non-maximal suppression•Hysteresis thresholdingThe Canny operatorGxGyGradient Vector: G = <Gx, Gy>Magnitude: |G| = sqrt(Gx2 + Gy2)ExampleOriginal ImageConvert toGrayscaleImage IntensityExample: Gx and GyGrayscale GyGxGyGxGradient MagnitudeExample: MagnitudeNon-maximal suppressionGyGxEdge Candidates(by Gx/Gy ratio)Hysteresis Thresholding•HIGH value:•LOW value:Magnitude must exceed HIGH to be considered an edgeNeighboring magnitudes must exceed LOW to be edge-worthyGradient MagnitudeEdge Candidates(by Gx/Gy ratio)EdgesThesholding
View Full Document