Font Size: a A A

Moment Technique And Its Applications In Image Processing And Recognition

Posted on:2003-05-14Degree:DoctorType:Dissertation
Country:ChinaCandidate:X H WangFull Text:PDF
GTID:1118360092466154Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
This paper aims at the research of moment technique and its applications in image processing and recognition. In general,moments describe numeric quantities at some distance from a reference point or axis. Moments are commonly used in statistics to characterize the distribution of random variables,and. similarly,in mechanics to characterize bodies by their spatial distribution of mass. The use of moments for image analysis is straightforward if we consider a binary or gray level image segment as a two-dimensional density distribution function. In this way,moments may be used to characterize an image segment and extract properties that have analogies in statistics and mechanics. In recent years,the excellent properties of moments and moment invariants computed from two-dimensional or three-dimensional shapes have aroused great attention,and researchers explored a wide range of moment applications. There are few image feature techniques that can be directly compared to the moment approach,so it is necessary to have a thorough research of moment technique and its applications in image processing and recognition. The research work of this paper can be classified in the following respects:1 In the aspect of moment techniques and properties,we mainly discuss the definition and properties of regular moment,orthogonal moments and other moments and invariants. We also compare and evaluate the performance of various moments in image representation,noise sensitivity,and information redundancy. For overall performance,Zernike moment and pseudo-Zernike moment are the best.2 In chapter three,we summarize the fast algorithms for regular moment and orthogonal moments and evaluate their performances. Computing the moment of two-dimensional images using straightforward algorithm needs lots of additions and multiplications. However,operation speed is crucial in many applications,especially in real-time pattern recognition applications. So it is necessary to investigate fast algorithms for reducing computation complexity.3 Chapter three puts forward an efficient fast algorithm for regular moment,which is named line segment method. Using this algorithm,we can get precise results of regular moments and moment invariants for arbitrary binary images. This aspect breaks through the limitation of Delta method,which suits only horizontal convex binary images. The experiments on the images of triangle,rectangle and plane approve the correctness and validity of line segment method.4 Chapter four is devoted to summarize moment applications in the fields of image processing,computer vision,and pattern recognition. These applications mainly include scene matching,histogram matching,image reconstruction,image compression,symmetry detection,image normalization,texture segmentation,edge detection,aircraft recognition,and imageindexing,etc..5 Chapter five provides a nev,method for image normalization,which can correctly normalize the images distorted by translation,scaling,rotation,and skew transformations,while extant normalization methods can normalize three distortions at most. This new method first process the input image using compact algorithm,then rotate the compact image according toimage eclipse tilt angle and the sign of u30. The experiment on plane and Chinese word images validate the correctness of this new image normalization method.6 In chapter six,we implement and analyze an optical character recognition system. This system uses features based on statistic moments to recognize uppercase and lowercase English characters. These moments include normalized central moments,Hu's moment invariants,affine moment invariants,and Tsirikolias-Mertzios moments. The classification techniques used here are Euclidean distance measure,normalized cross correlation and discrimination cost. The mean of the intraclass standard deviations of the features is used as a weighting factor during the classification process to improve recognition accuracy. The system is rigorously tested under different conditions,including using diff...
Keywords/Search Tags:regular moment, moment invariant, orthogonal moment, line segment method, image normalization, image eclipse tilt angle, optical character recognition, combined invariant, weighted normalized cross correlation
PDF Full Text Request
Related items