Towards Feature Detection based on Morphology of Objects on Image

Description
Title: Towards Feature Detection based on Morphology of Objects on Image
Authors: Solis Montero, Andres
Date: 2010
Abstract: This thesis describes a new line segment detection and extraction algorithm for computer vision, image segmentation, and shape recognition applications. This algorithm uses a compilation of different image processing techniques such as normalization, Gaussian smooth, automatic threshold, and Laplace edge detection to extract edge contours from color input images. Contours of each connected component are divided into short segments, which are classified by their orientation into about ten discrete categories. Straight lines are recognized as the minimal number of such consecutive short segments with the same direction. This solution indeed gives us more precise line segments (including line endpoints) and requires a shorter time than the widely used Hough Transform algorithm for detecting line segments given any orientation and location inside an image. Its easy implementation, simplicity, parameter minimization, speed, ability to split an edge into straight line segments using the actual morphology of objects, accuracy and the use of OpenCV library are key features and advantages of the proposed approach. The algorithm was tested on several simple shape images as well as on real pictures, yielding a more accurate resemblance of straight lines in accordance with the human perception of line taxonomy. The line detection algorithm introduced here requires few parameters and is robust to standard image transformations such as rotation, scaling and translation. Furthermore, some of these parameters are selected by automatic unsupervised methods, thus improving the expected algorithm outcome in terms of the stated problem. Several experimental results are presented to support the validity of the algorithm.
URL: http://hdl.handle.net/10393/28558
http://dx.doi.org/10.20381/ruor-19332
CollectionTh├Ęses, 1910 - 2010 // Theses, 1910 - 2010
Files
MR65976.PDF4.6 MBAdobe PDFOpen