Quaternion Based Full Reference Color Image Quality Assessment Methods And Their Extension

With more digital information appeared,various functional systems have emerged for image,image acquisition,processing,etc.However,the imperfections of imaging systems,transmission media and storage devices,with object motion and noise pollution causes image distortion and degradation,which is inconvenient for subsequent processing.Therefore,it is very important to evaluate the image quality reasonably.This paper firstly uses quaternion to represent color image,designing two evaluation indicators considering with visual saliency and structural feature.In addition,extended the full reference image quality evaluation to no reference image quality assessment by Broad Learning System.The main work of the thesis is as follows:(1)A Quaternion Spectral Residual SIMilarity(QSR-SIM)method is proposed.Firstly,the reference image and the distorted image are represented as a pure quaternion matrix,using the quaternion spectral residual method to detect the visually significant region of the two images;Secondly,merging the quaternion gradient features and used as two indicators of Color Image Quality Assessment(CIQA).Visual saliency regard as the weight of the evaluation index.Finally,Numerical experiments were carried out in two standard image quality evaluation databases.The results show that die new method is consistent with the subjective evaluation of humans.(2)The Quaternion Gradient Magoitude Similarity Deviation(QGMSD)method is proposed.Firstly,measuring the local similarity quality score of the color image gradient.Secondly,using the similarity deviation to reflect the global variation of the distorted image stnicture.The results show that the algorithm is competitive.(3)A Broad Learning System for Image Quality Assessment(BLS-IQA)is proposed.Firstly,extracting the gradient and the distribution characteristics of Mean Subtracted Contrast Normalized.Then,the general non-reference image quality evaluation method is obtained by using the Broad Learning System.The experimental results show that the method is highly consistent with the subjective evaluation of human.