Books Edge Detection Based On Digital Image Technology

It is known that digital image’s edge detection and boundary tracing is always an important and difficult task of image processing. In this thesis, Based on analyzing and comparing several typical edge detection algorithm, I selected the corresponding edge detection and boundary tracing methods, and according to the characteristics of book images, made some improvement program, realized the boundary tracing of the curve in book image perfectly.In the book image’s edge extraction process, when using the Otsu method for image binarization processing, I used a combination of local and global, dynamic threshold method, to make the edge extraction process of binarized image more effective.By analyzing a variety of edge detection operator, such as Roberts, Canny, this thesis chose to use LOG (Laplace of Gauss) operator, and according to the shape of book content normally fixed to be rectangular, and we extracted edge side by side, so using a second derivative of the one-dimensional convolution method similar template in the vertical direction and the horizontal direction of the pixels separately, minimized the amount of calculation without reducing the detection effect.In the process of boundary tracing, combined graph theory method and the dynamic programming algorithm effectively in this thesis, then searching the minimum cost path to detect image edge, ensuring the boundary tracking to be uninterruptedly and accurately.In addition, in consideration of high image quality, large amount of computation of the situation, I put forward the idea’coarse-to-fine’ phased detection methods. It took significant effect for improving performance obviously.Experimental results show that the proposed solution in this thesis can accurately and efficiently detect and track the book image’s whole boundary, on the premise of regular size, backgrounds mainly black book images.