Study On Degraded Image Invariant Recognition

Degraded image invariant recognition is an important theory to improve the adaptability of computer vision system. Image degradation can be divided into two classes, image geometric structure based degradation and image gray level based degradation. Image geometric based degradation includes image affine transform (i.e. translation, scale, rotation, clip and so on.) and the nonlinear distortion of the camera. Image gray level based degradation includes various noises, out-focus blur, motion blur, atmosphere turbulence and diffusion, severe weather (i.e. fog, haze and sand storm), image color change (illumination, contrast, saturation). In this paper, we studied the degraded image invariant recognition problem in term of image global feature methods, such as moments based image analysis and image transforms. The mainly methods used are list as followed1. A new set of moments based on the Bessel function of the first kind, named Bessel-Fourier moments is proposed in this paper. The new proposed moments are more suitable than Orthogonal Fourier-Mellin and Zernike moments for image analysis and rotation invariant pattern recognition. Theoretical and Experimental results show that the Bessel-Fourier Moments perform better than the Orthogonal Fourier-Mellin and Zernike moments (OFMMs and ZMs) in terms of image reconstruction capability and invariant recognition accuracy in noise-free, noisy and smooth distortion conditions. In addition, the Bessel-Fourier moments can be thought of as generalized orthogonalized complex moments. Therefore, the feast computation can be easily realized by computing complex moments in the Certesian coordinate system directly.2. A new method to construct a set of combined blur, translation, scale and rotation invariant features using Radon and Fourier-Mellin transforms is proposed in this paper, named Radon and pseudo Fourier-Mellin invariants (RPFMI). The proposed method is robust to additive white noise as a result of summing pixel values to generate projections in the Radon transform step. We also present a mathematical framework of obtaining the Radon and pseudo Fourier-Mellin transforms of blurred images, and a framework of deriving the combined blur, scale and rotation invariants. Theoretical and experimental results show the superiority of the proposed method and its robustness to additive white noise in comparison with some recent methods.3. Radial Tchebichef moments as a discrete orthogonal moment in the polar coordinate have been successfully used in the field of pattern recognition. However, the scaling invariant property of these moments has not been studied due to the complexity of the problem. Image normalization was usually used to achieve the scaling invariant. The recognition accuracy was affected since normalization scheme leads to the loss of some characteristics of an image. In this paper, we present a new method to construct a complete set of scaling and rotation invariants extract from radial Tchebichef moments, named radial Tchebichef moment invariants. Experimental results show the efficiency and the robustness to noise of the proposed method for recognition tasks.4. A generic approach based on Jacobi-Fourier moments for scale and rotation invariant analysis of radial orthogonal moments is also proposed in this paper. It provides a fundamental mathematical tool for invariant analysis of the radial orthogonal moments since Jacobi-Fourier moments are the generic expressions of radial orthogonal moments. Zernike Moment invariants, Orthogonal Fourier-Mellin Moment invariants, pseudo-Zernike Moment invariants and radial Tchebichef moment invariants all are the variation of Jacobi-Fourier moment invariants. Theoretical and experimental results show the description performance and the robustness to noise of the proposed method.5. A novel generic method for circularity measure is proposed in this paper. The proposed method is based on polar radius moment and satisfies the desirable properties of circularity measure. Compared with the most standard circularity measure, the new measure is easy to compute, is robust to noise and performs better in the case of shapes with boundary defects. Theoretical and experimental results also show the superiority of the proposed method. In addition, the proposed method is a generalized measure so that shape circularity value can be easy controlled by alternating the parameter p. The influence of parameter p on shape circularity value and a suitable choice of the parameter are also given in this paper.