M.A. Thesis:

A new edge detection filter based on Sobel and Prewitt filters

One of the most important operations in the segmentation of image editing is edge detection. The purpose of the edge detection is to extract the important features of the image so that it can be interpreted correctly from the image. The edges are parts of the image in which the intensity of the color changes between the two points of the image called the pixel. So far, there are many ways in which edge detection is provided, each with advantages and disadvantages. According to the definition of the edge that is derived in equivalence mathematics, these methods are divided into two general categories, which are: first derivative methods (gradients): in these methods, the edge is where the first derivative has the highest value, such as methods Sobel, Prewitt, Roberts, etc. These methods have high execution speeds and low computing but have a high sensitivity to noise and, in many cases, do not preserve the bonding of the edges. The second group is the second-derivative methods, in which these methods are the edge where the second derivative is zero, such as the Laplacian method.These methods have a lower sensitivity to noise and provide better visualization than edge gradient methods but have more computation. Since the correct identification of the edge in various images is still one of the challenges that has not been completely solved, in this research, we tried to use a traditional image-processing concept called mathematical morphology using a traditional edge detection method and its combination with one of the important concepts of image processing. Can better identify the edges of an image than the described methods. The proposed method was implemented in MATLAB simulation environment and the results were compared using charts and tables with other traditional methods. To compare the method Proposed traditional image quality parameters such as PSNR, SSIM and SI are used Comparison of proposed method with three traditional methods shows that according to the values of the PSNR and SSIM parameters, the proposed method more accurately detects the edges of the image and and It has better quality according to the values of the SI parameter , the proposed method is less clear than traditional methods and cannot be guaranteed to be superior to traditional methods

Keywords: : Edge detection-Gradient-Laplacein-Morphology-Image Quality Parameters