Presentation is loading. Please wait.

Presentation is loading. Please wait.

EDGE DETECTION Instructor: Dr. Engr. Junaid Zafar.

Similar presentations


Presentation on theme: "EDGE DETECTION Instructor: Dr. Engr. Junaid Zafar."— Presentation transcript:

1 EDGE DETECTION Instructor: Dr. Engr. Junaid Zafar

2 2 CONTENTS  Introduction  Types of Edges  Steps in Edge Detection  Methods of Edge Detection First Order Derivative Methods  First Order Derivative Methods - Summary Second Order Derivative Methods  Second Order Derivative Methods - Summary Optimal Edge Detectors  Canny Edge Detection  Edge Detector Performance  Line Detection Convolution based technique Hough transform  Application areas

3 Edges (or Edge points) are pixels at or around which the image values undergo a sharp variation. Edge Detection: Given an image corrupted by acquisition noise, locate the edges most likely to be generated by scene elements, not by noise.

4 Edge Formation: 1.Occluding Contours 1.Two regions are images of two different surfaces 2.Discontinuity in surface orientation or reflectance properties

5 Mathematical Model of edges and noise

6 More Realistically, due to blurring and noise, we generally have

7 Image Gradients The gradient of a differentiable function I gives the direction in which the values of the function change most rapidly. Its magnitude gives you the rate of change.

8 Approximating Derivatives

9 Local Operators (Differential Operators) accentuate noise !! (why?) Therefore, need smoothing before computing image gradients. Motivation: Smoothing removes local intensity variation and what remains are the prominent edges. Gaussian Smoothing with exponential kernel function

10 Three Steps of Edge Detection 1.Noise Smoothing: Suppress as much of the image noise as possible. In the absence of specific information, assume the noise white and Gaussian 2.Edge Enhancement: Designe a filter responding to edges. The filter’s output is large at edge pixels and low elsewhere. Edges can be located as the local maxima in the filters’ output. 3.Edge Localization: Decide which local maxima in the filter’s output are edges and which are just caused by noise. 1.Thinning wide edges to 1-pixel with (nonmaximum suppression); 2. Establishing the minimum value to declare a local maximum an edge ( thresholding )

11 Edge Descriptors (The output of an edge detector) 1.Edge normal: The direction of the maximum intensity variation at the edge point. 2.Edge direction: The direction tangent to the edge. 3.Edge Position : The location of the edge in image 4.Edge strength: A measure of local image contrast. How significance the intensity variation is across the edge.

12

13 13 Steps in Edge Detection  Edge - Area of significant change in the image intensity / contrast  Edge Detection – Locating areas with strong intensity contrasts  Use of Edge Detection – Extracting information about the image. E.g. location of objects present in the image, their shape, size, image sharpening and enhancement

14 14 TYPES OF EDGES  Variation of Intensity / Gray Level Step Edge Ramp Edge Line Edge Roof Edge

15 15 Steps in Edge Detection  Filtering – Filter image to improve performance of the Edge Detector wrt noise  Enhancement – Emphasize pixels having significant change in local intensity  Detection – Identify edges - thresholding  Localization – Locate the edge accurately, estimate edge orientation

16 16 METHODS OF EDGE DETECTION  First Order Derivative / Gradient Methods Roberts Operator Sobel Operator Prewitt Operator  Second Order Derivative Laplacian Laplacian of Gaussian Difference of Gaussian  Optimal Edge Detection Canny Edge Detection

17 17 First Derivative  At the point of greatest slope, the first derivative has maximum value E.g. For a Continuous 1- dimensional function f(t)

18 18 Gradient  For a continuous two dimensional function Gradient is defined as

19 19 Gradient  Approximation of Gradient for a discrete two dimensional function  Convolution Mask Gx = Gy =  Differences are computed at the interpolated points [i, j+1/2] and [i+1/2, j] 1 1 11 1 1

20 EE - 6358 Computer Vision20 Gradient Methods – Roberts Operator  Provides an approximation to the gradient  Convolution Mask Gx= Gy =  Differences are computed at the interpolated points [i+1/2, j+1/2] and not [i, j] 1 0 0 0 10

21 21 Roberts Operator - Example  The output image has been scaled by a factor of 5  Spurious dots indicate that the operator is susceptible to noise

22 22 Gradient Methods – Sobel Operator  The 3X3 convolution mask smoothes the image by some amount, hence it is less susceptible to noise. But it produces thicker edges. So edge localization is poor  Convolution Mask Gx = Gy=  The differences are calculated at the center pixel of the mask. 01 -202 01 121 000 -2

23 23 Sobel Operator - Example  Compare the output of the Sobel Operator with that of the Roberts Operator: The spurious edges are still present but they are relatively less intense compared to genuine lines Roberts operator has missed a few edges Sobel operator detects thicker edges Will become more clear with the final demo Outputs of Sobel (top) and Roberts operator

24 24 Gradient Methods – Prewitt Operator  It is similar to the Sobel operator but uses slightly different masks  Convolution Mask Px = Py = 01 01 01 111 000

25 25 First Order Derivative Methods - Summary  Noise – simple edge detectors are affected by noise – filters can be used to reduce noise  Edge Thickness – Edge is several pixels wide for Sobel operator– edge is not localized properly  Roberts operator is very sensitive to noise  Sobel operator goes for averaging and emphasizes on the pixel closer to the center of the mask. It is less affected by noise and is one of the most popular Edge Detectors.

26 26 Second Order Derivative Methods  Zero crossing of the second derivative of a function indicates the presence of a maxima

27 27 Second Order Derivative Methods - Laplacian  Defined as  Mask  Very susceptible to noise, filtering required, use Laplacian of Gaussian 010 1-41 010

28 28 Second Order Derivative Methods - Laplacian of Gaussian  Also called Marr-Hildreth Edge Detector  Steps Smooth the image using Gaussian filter Enhance the edges using Laplacian operator Zero crossings denote the edge location Use linear interpolation to determine the sub-pixel location of the edge

29 29 Laplacian of Gaussian – contd.  Defined as  Greater the value of , broader is the Gaussian filter, more is the smoothing  Too much smoothing may make the detection of edges difficult

30 30 Laplacian of Gaussian - contd.  Also called the Mexican Hat operator

31 31 Laplacian of Gaussian – contd.  Mask Discrete approximation to LoG function with Gaussian = 1.4

32 32 Second Order Derivative Methods - Difference of Gaussian - DoG  LoG requires large computation time for a large edge detector mask  To reduce computational requirements, approximate the LoG by the difference of two LoG – the DoG

33 33 Difference of Gaussian – contd.  Advantage of DoG Close approximation of LoG Less computation effort Width of edge can be adjusted by changing 1 and 2

34 34 Second Order Derivative Methods - Summary  Second Order Derivative methods especially Laplacian, are very sensitive to noise  Probability of false and missing edges remain  Localization is better than Gradient Operators

35 35 Optimal Edge Detector  Optimal edge detector depending on Low error rate – edges should not be missed and there must not be spurious responses Localization – distance between points marked by the detector and the actual center of the edge should be minimum Response – Only one response to a single edge  One dimensional formulation Assume that 2D images have constant cross section in some direction

36 36 Mathematical Formulation of Performance Criteria  Notation Edge : G(x) Noise : n(x) Impulse response of Filter : f(x)  Assumptions The edge is located at x=0 Filter has finite response bounded by [-W, W] Noise is of Gaussian nature n 0 : Root mean squared noise amplitude per unit length

37 37 First Criterion: Edge Detection  Response of filter to the edge:  RMS response of filter to noise :  First criterion: output Signal to Noise Ratio

38 38 Second Criterion: Edge Localization  A measure that increases as the localization increases is needed  Reciprocal of RMS distance of the marked edge from center of true edge is taken as the measure of localization  Localization is defined as:

39 39 Maximizing the product  The design problem is reduced to maximization of the product:

40 40 Third Criterion – Elimination of multiple responses  In presence of noise several maxima are detected – it is difficult to separate noise from edge  We try to obtain an expression for the distance between adjacent noise peaks  The mean distance between the adjacent maxima in the output is twice the distance between the adjacent zero crossings in the derivative of output operator

41 41 Elimination of Multiple Responses – contd.  Expected number of noise maxima in the region of width 2W is given by  Fixing k will restrict the number of noise maxima that could lead to false response – this is a constraint imposed on the design

42 42 Finding Optimal Detectors by Numerical Optimization  It is not possible to find a function f which will maximize the SNR, Localization product in the presence of the multiple response constraint  If the function f is discrete then the computation is simplified to four inner products  Additional constraints can be included by the use of Penalty Function – it has non-zero value when the constraint is violated  f is then found out for maximizing the following expression

43 43 Detector for Step Edges  G(x) = Au -1 (x)

44 44 Spatial scaling  We can scale the filter width  Detection improves with width, localization degrades with increase in width  The uncertainty principle is

45 45 Efficient approximation  Depending on the above principles, several optimal edge detectors are calculated  Best approximation to the above detectors is the First Derivative of Gaussian  It is chosen because of the ease of computation in 2 dimensions

46 46 Noise estimation  Important to estimate the amount of noise in the image to set thresholds  Noise component can be efficiently isolated using Weiner Filtering – requires the knowledge of the autocorrelation of individual components and their cross- correlation  Noise strength is estimated by Global Histogram Estimation

47 47 Thresholding  Broken edges due to fluctuation of operator output above and below the threshold – results in Streaking  Use double thresholding to eliminate streaking

48 48 Two Dimensional Edge Detection  In two dimensions edge has both position and direction  A 2-D mask is created by convolving a linear edge detection function aligned normal to the edge direction with a projection function parallel the edge direction  Projection function is Gaussian with same deviation as the detection function  The image is convolved with a symmetric 2-D Gaussian and then differentiated normal to the edge direction

49 49 Need for multiple widths and Feature Synthesis  Width is chosen to give best tradeoff between detection and localization  For different edges, width has to be different since SNR is different  Feature Synthesis Smaller operator responses are used to predict the larger operator response New edge points are marked if the actual large operator output varies significantly from the predicted output

50 50 Implementation of Canny Edge Detector  Step 1 Noise is filtered out – usually a Gaussian filter is used Width is chosen carefully  Step 2 Edge strength is found out by taking the gradient of the image A Roberts mask or a Sobel mask can be used

51 51 Implementation of Canny Edge Detector – contd.  Step 3 Find the edge direction  Step 4 Resolve edge direction

52 52 Canny Edge Detector – contd.  Step 5 Non-maxima suppression – trace along the edge direction and suppress any pixel value not considered to be an edge. Gives a thin line for edge  Step 6 Use double / hysterisis thresholding to eliminate streaking

53 53 Canny Edge Detector – contd.  Compare the results of Sobel and Canny

54 54 Edge Localization – Subpixel Location Estimation  Take Gaussian edge detector output without non-maxima suppression  Error displacement is computed by taking the weighted mean of distance, along the edge gradient on either side of the edge

55 55 Edge Detector Performance  Criteria Probability of false edges Probability of missing edges Error in estimation of edge angle Mean square distance of edge estimate from true edge Tolerance to distorted edges and other features such as corners and junctions

56 56 Figure of Merit  Basic errors in a Edge Detector Missing Edges Error in localizing Classification of noise as Edge I A : detected edges I I : ideal edges d : distance between actual and ideal edges  : penalty factor for displaced edges

57 57 Line Detection  Lines in an image present another form of discontinuity  It is worthwhile to consider them separately since they are very frequent  Methods Use of Convolution Mask Hough Transform

58 58 Line Detection using Convolution Mask  We need to decide the orientation and width of line to be detected  Following masks are for single pixel wide lines in four directions

59 59 Line Detection using Convolution Mask – contd.  The mask for horizontal orientation is applied  Lines are 5 pixels wide, mask is tuned for detecting 1 pixel wide line  Line detector acts as edge detector

60 60 Line detection – Hough Transform  Method to isolate the shapes from an image  Performed after edge detection  Not affected by noise or gaps in the edges  Technique Thresholding is used to isolate pixels with strong edge gradient Parametric equation of straight line is used to map the edge points to the Hough parameter space Points of intersection in the Hough parameter space gives the equation of line on actual image

61 61 Hough Transform - contd  Parametric Equation of Straight Line

62 62 Hough Transform – contd. Edge Detected Image Mapping of Edge Points in Hough Parameter Space

63 63 Applications  Enhancement of noisy images – satellite images, x-rays, medical images like cat scans  Text detection  Mapping of roads  Video surveillance, etc.

64 64 Applications  Canny Edge Detector for Remote Sensing Images Reasons to go for Canny Edge Detector – Remote sensed images are inherently noisy Other edge detectors are very sensitive to noise

65 65 Edge map of remote sensed image using Canny

66 66 Image enhancement and sharpening using canny

67 Canny Edge Detector Original Image Edges

68 68 MATLAB Demo  Get an image from the internet  Apply following operators on the image and display the results Roberts Sobel Prewitt Laplacian of Gaussian Canny


Download ppt "EDGE DETECTION Instructor: Dr. Engr. Junaid Zafar."

Similar presentations


Ads by Google