Research On Image Quality Assessment Algorithm Based On Quaternion Wavelet Transform

Images as an important information carrier has become increasingly integrated into our daily life, and is also the key research direction in the computer vision and image processing. In practical applications, the original data or uncompressed image is difficult to obtain.we cannot effectively measure the current image quality, meanwhile, it greatly limits the application of the full reference and the reduce reference image quality assessment methods. Therefore, it particularly important to develop a simple and effective blind(no referen ce) image evaluation algorithm.Quaternion wavelet transform is developed on the basis of dual tree complex wavelet transform, which overcomes malpractice in traditional wavelet transform, such as aliasing properties, Aliasing sex, lack of translation invariance and directional defects. Because of the multi-scale analysis advantages and less redundancy, QWT amplitude and phase information has unique advantages in two-dimensional signal feature extraction. C haracteristics of each sub-band can be used in the amplitude, at the same time, phase provides local features. As more image texture features information are selected, and in some extent QWT provides a good foundation for supervised machine learning classification algorithm.In this paper,we mainly studies the blind reference image quality assessment algorithm based on the QWT transform.the relative research contents and results are as follows:1. A new general purpose no-reference image quality assessment based on quaternion wavelet transform is proposed. It projects a two dimensional image to four dimensional space by using the quaternion wavelet pyramid. Each level of pyramid provides a shift- invariant magnitude and 3-angle(?,?, ?) phase. As phase ?includes image texture information and can effectively characte rize the structural information of the image. So the method extracts the features which can reflect the degree of image distortion and constitute the feature vector.Finally, the extracted feature is inputted to the SVR(Support Vector Regression, SVR) model to predict image quality score.Experimental result shows that the This algorithm can effectively reflect the visual quality of the image in different distortion types. And Spearman correlation coefficient(SROCC) value can reaches 0.942.2. Stereoscopic image quality assessment algorithm via multi- core learning. Study the image sub-band modeling and wavelet coefficient distribution model parameter estimation problem through generalized Gaussian Distribution.Firstly,the algorithm utilizing the generalized Gaussian distribution fits the sub-band coefficient.Meanwhile,we take into account the depth information in the stereoscopic image.finally combined with fitting parameters to measure the degree of distortion, and then obtain the objective assessment of the distorted image.experimental results show that the algorithm is suitable for all kinds of image distortion, and has good consistency with subjective evaluation.3.The paper proposed a new no-reference hybrid distortion image quality assessment algorithm based on natural scene statistics through in the spatial and frequency domain, The original image is not only in the spatial domain with a certain distribution, but also follow certain distribution in the frequency domain. however, the distortion will change its distribution characteristics,and then o would lead to have a certain bias in Statistical parameters of images.the method measure distorted image and the original image bias which can be reflect in the statistical law in spatia l and frequency domain. Experimental results show that the algorithm is superior to the traditional objective evaluation algorithm, and has a good consistency with subjective human perception.