Font Size: a A A

Study On Methods For Color Texture Image Analysis Using Fractal Geometry

Posted on:2016-02-18Degree:DoctorType:Dissertation
Country:ChinaCandidate:F ZhuFull Text:PDF
GTID:1108330482975123Subject:Image Processing and Scientific Visualization
Abstract/Summary:PDF Full Text Request
Texture image analysis is an important research area in the field of image processing and computer vision. It has a very wide range of applications in agricultural products detection, remote sensing image analysis and intelligent transportation systems. Texture analysis technology is the process of understanding and recognition of natural texture image by imitating human vision. Texture feature description and extraction is an important basis for texture analysis. Fractal objects exist widely in nature, such as the curling up smoke, different poses clouds, a lush growth of trees and rough surface. In fact, the fractal object is also a texture. Therefore, there is a natural connection between fractal and texture, which takes the fractal theory as an important method in the field of image processing. However, current methods of texture image analysis based on fractal theory can only cope with binary images or gray images, while most natural textures are color images. Therefore, it is still an open and challenging problem to propose an effective, reliable and being of human visual pattern method to analyse the color texture images.In order to overcome some disadvantages of current color natural texture analysis methods, this paper proposed some methods of texture feature expression, feature extraction and image classification using several kinds of color fractal dimension of natural images. Research works of this paper are shown as following:(1) As the traditional lexicographical ordering can not effectively use every component information of the vector, we presented a novel vector extra method based on clustering and extended gray scale morphology to vector morphology. Then, we use the dilation operation of vector morphology in HSI color space to calculate color image fractal dimension directly replacing box-counting commonly. Moreover, the union features, which are made of the color fractal feature and the other color channel fractal features, are defined to research color image classification problem with multi-classes. Finally, the experimental results show that our proposed approach is more effective than other color texture analysis methods in two aspects of correct classification rate and time complexity.(2) Fractal descriptors based on quaternion Fourier transform and local polynomial regression was proposed for color texture image analysis. Firstly, considering the relation between the power spectrum and frequency in the quaternion Fourier transform domain, we proposed fractal dimensions of a color image using a fast quaternion Fourier transform. Second, local polynomial regression is applied to estimate log(spectrum)-log(frequency) curve, which is not usually linear in a natural texture image. Finally, a local polynomial regression curve is defined as fractal descriptors for the color image classification problem with multi-classes. The experimental results show that our proposed approach is more effective than other color texture analysis methods both in the correct classification rate and the duration.(3) A color texture feature extraction method based on quaternion Gabor wavelet transform and multifractal dimension was presented. Firstly, considering the non-commutative property of the quaternion multiplication, a quaternion is decomposed into two complex numbers before the quaternion Fourier fast transform algorithm is used for quaternion Gabor wavelet transform. Then the generalized fractal dimension of each filtered image is extracted as a color texture feature vector, where the amplitude of each pixel is used as fractal surface quality. Finally, the application of color image classification to verify the effectiveness of proposed method.(4) According to the existing local connection fractal dimension can not be used for color image but only for binary image or gray image, a novel local connection fractal dimension of color image was proposed. Firstly, the color value of each pixel and the coordinate of the position are made of five dimensional vectors. Secondly, the distance between two points is used to determine the connectivity between pixels and to create a graph model. Then the local fractal dimension is defined as the local connectivity, which is the number of vertex of the minimal connected sub-graph and to obtain each pixel. Finally, the local fractal dimension is applied in the segmentation of blood vessel image by a series of processes such as image denoising, image enhancement and image binary.
Keywords/Search Tags:Fractal Dimension, Texture Analysis, Vectorial Morphology, Quaternion Fourier Transform, Local Polynomial Regression, Quaternion Gabor Transform, Multi -fractal, Local Connection Fractal Dimension, Blood Vessel Image Segmentation
PDF Full Text Request
Related items