Edge Detection in Image Processing
EDGE DET ECT I ON I N I M AGE PROCESSI N G
CANNY’S EDGE DETECTOR
What are edges in an image?
Input Image Output Image
Look for places in the image where Intensity changes quickly
Edges are pixels where image brightness changes abruptly
Significance of Edge Detection
Significantly reduces the amount of data to be processed and filters out useless info, while preserving the imp. features
Steps in Canny’s Edge Detector
Image Smoothening Finding the gradient Non-maximum suppression Image Thresholding
Image Smoothening (blurring)
Original Image After Smoothening
•The image is convolved with gaussian filter •High frequency noise is eliminated
Finding the gradient at every pixel
Intensity value at every pixel is available :I(x,y) Calculate gradient magnitude (done by convolving the image with Sobel operator) Calculate gradient directions
Finding Gradient (Conti…)
Calculates gradient Gx and Gy Total gradient: sqrt(Gx2 + Gy2) Direction: invtan(Gy/Gx) Gradient direction is Quantized: horizontal, vertical and diagonal
Non- maximum suppression
Let M(x,y) be the gradient in a given direction M(x,y) > M(x+dx, y+dy) M(x,y) > M(x-dx, y-dy) Condition not satisfied- suppress that pixel (cannot act as an edge)
Thresholding (eliminating unwanted edges)
2 thresholds : tlow and thigh If M(x,y) < tlow ? This pixel cannot represent an edge If M(x,y) > thigh ? This pixel represents an edge If tlow < M(x,y) < thigh ? check for connectivity with an edge pixel
Thank you…
Original Image After applying Canny’s Edge detector
CANNY’S EDGE DETECTOR
What are edges in an image?
Input Image Output Image
Look for places in the image where Intensity changes quickly
Edges are pixels where image brightness changes abruptly
Significance of Edge Detection
Significantly reduces the amount of data to be processed and filters out useless info, while preserving the imp. features
Steps in Canny’s Edge Detector
Image Smoothening Finding the gradient Non-maximum suppression Image Thresholding
Image Smoothening (blurring)
Original Image After Smoothening
•The image is convolved with gaussian filter •High frequency noise is eliminated
Finding the gradient at every pixel
Intensity value at every pixel is available :I(x,y) Calculate gradient magnitude (done by convolving the image with Sobel operator) Calculate gradient directions
Finding Gradient (Conti…)
Calculates gradient Gx and Gy Total gradient: sqrt(Gx2 + Gy2) Direction: invtan(Gy/Gx) Gradient direction is Quantized: horizontal, vertical and diagonal
Non- maximum suppression
Let M(x,y) be the gradient in a given direction M(x,y) > M(x+dx, y+dy) M(x,y) > M(x-dx, y-dy) Condition not satisfied- suppress that pixel (cannot act as an edge)
Thresholding (eliminating unwanted edges)
2 thresholds : tlow and thigh If M(x,y) < tlow ? This pixel cannot represent an edge If M(x,y) > thigh ? This pixel represents an edge If tlow < M(x,y) < thigh ? check for connectivity with an edge pixel
Thank you…
Original Image After applying Canny’s Edge detector