Studies On Inclined Plane Decomposition Based Non-symmetry And Anti-packing Image Representation Methods And Processing Algorithms

An image system model consists of two parts: image representation and imageprocessing. Image representation is the manner in which images are described and storedin the computer. Image representation plays a key role in optimizing image processingalgorithms. Selection of different image representation methods always determines theefficiency of image processing algorithms to a great extent. Based on the Non-symmetryAnti-packing pattern representation Model (NAM), a new image representation method ispresented, which is called Inclined plane Decomposition based Non-symmetry andAnti-packing image representation Method (IDNAM), in order to improve the efficiencyof image processing algorithms. A series of instances of rectangular inclined planesub-patterns are used to represent an image in IDNAM. A grey scale image representationmethod is proposed based on the distorted IDNAM. A homogeneity determination rule fora rectangular region of a grey scale image is introduced to limit the distortion caused byanti-packing in IDNAM. Experimental results show that the IDNAM representation forgrey scale images can not only extract rectangular inclined plane sub-patterns withdifferent sizes according to the image contents, but also keep a high fidelity. A binaryimage representation method is proposed based on the undistorted IDNAM. An algorithmis given for an abstract image processing based on IDNAM, and the algorithm sums up theentire idea of image processing algorithms based on IDNAM. It is a general, abstract andguide algorithm. The theoretical analyses show that the time complexity of this algorithmis smaller than that of the abstract image processing algorithms based on pixelrepresentation, when rectangular inclined plane sub-patterns are less than pixels in animage.Integral projection is a basic method in image analysis, which is widely used inseveral domains such as face recognition, line detection and skew detection etc. Thetraditional integral projection algorithm is developed using the pixel representation. Bystudying the method of computing the integral projection vector of a rectangular inclinedplane sub-pattern and the relationship between the integral projection vectors of arectangular inclined plane sub-pattern and an image, a fast integral projection algorithm isproposed, which achieves computing the integral projection vectorof an image fromrectangular inclined plane sub-patterns directly. The theoretical analyses and experimentalresults show that significant improvement is obtained with the algorithm using theIDNAM representation over the pixel representation along all projection directions. Edge detection, which measures, checks and captures discontinuity in gray levelvalues in an image, is a basic problem in image analysis. Edge detection, as an initial step,is widely used in several domains such as image segmentation, pattern recognition andmotion detection etc. Classical edge detection algorithms are based on pixel representation,which extract edge information from pixels. On the basis of the IDNAM representation forgrey scale images, an ideal edge model within a rectangular inclined plane sub-pattern ispresented. In the model, edges within rectangular inclined plane sub-patterns are dividedinto five types: central edge model, left border edge model, top border edge model, rightborder edge model and bottom border edge model. The methods for computing the edgestrength and direction of the five edge models are presented. The existence of an edgewithin a rectangular inclined plane sub-pattern should be determined according to theparameters of the five edge models, to separate between edge and noise. The theoreticalanalyses and experimental results show that the edge detection algorithm using IDNAMperforms faster than the classical ones because it permits the execution of operations onsub-patterns instead of pixels.Segmentation of an image entails the division or separation of the image into regionswith similar attribute. Image segmentation is a common technology in many applicationsof image analysis and computer vision. A fast image segmentation algorithm is proposedbased on IDNAM. According to the edge strength within rectangular inclined planesub-patterns and the homogeneity between them, the algorithm groups the rectangularinclined plane sub-patterns into different regions, to achieve the segmentation. Thetheoretical analyses and experimental results show that the image segmentation methodusing IDNAM performs faster than the watershed algorithm.The theoretical analyses and experimental results show that the IDNAM imagerepresentation method can speedup image processing effectively.