Algorithms for Edge Detection CSE Seminar Report

Introduction to  Algorithms for Edge Detection Seminar:

As they say a picture speaks for a thousand words, a nice picture should express itself in the best possible way. There are so many softwares available in the market for image editing and processing. Most of us look at the beauty side of it but experts look for a perfect image with sharp and accurate edges with no discontinuities.

Sometimes there would be many irregularities in the boundary or edges of the image and in order to identify this type of irregularity edge detection process is available. This is done with the help of various edge operators which works for edge specific purposes. This operator is basically a 2-D filter which performs operation on the image edges and show null value at uniform edges and detects any type of large gradients within the image. There are many variables to determine which operator would be apt for a particular edge. Some of them are simple things by which we can decide the operator.

 Steps involved: 

  • Orientation of the edges
  • Detection of noisy images 

The structures of the edges also come into play as a gradual change in refraction may lead to non-unified edges. 

Edge detection processes are classified into: 

  • Gradient category
  • Laplacian category 

Gradient category mainly deals with the image derivation and often is calculated on the average basis of the same whereas Laplacian method deals with ramp’s I-D shape. The zero crossings in the second derivation are searched for any edges. 

There are many edge detection techniques such as Sobel operator which constitutes convolution kernel pairs; Robert’s cross operator which estimates spatial gradient’s absolute magnitude, Prewitt’s operator detects both horizontal and vertical edges, Laplacian of Gaussian which seals with second derivation 2-D isotropic measure, Canny’s edge detection algorithm which is more like a guideline for edge detection process.

Download   Algorithms for Edge Detection CSE Seminar Report.

Leave a Reply

Your email address will not be published. Required fields are marked *