No-Reference Image Quality Assessment Based On PLSA

Images have been throughout all areas of human’s life and work in the information-digitalized times. It is of great significance to construct effective system for image quality assessment in the field of image processing and applications. No-reference image assessment not only meet the objective needs, but also has the possibility to realize.Thus,more and more attention has been paid to no-reference image assessment by many related research personnel.The main basis for the algorithm design of no-reference image quality assessment is statistical characteristics of images.But the research on the statistical characteristics of images stays in the lower stage,and the existing models are still too simple.The research shows that:human cognition especially vision cognition is hierarchical.Calculation problems about vision cognition can be solved by computer learning which can study the implicit factors in the data and build a corresponding model.Leading computer learning into the no-reference image quality assessment algorithm is helpful to an in-depth study on the natural statistical properties of images.Probabilistic latent semantic analysis (PLSA) is a typical topic model of the "unsupervised" machine learning.It was used for text classification initially,but it has been widely used in the hot research in computer vision study including scene classification and behavior analysis at present.This paper introduces PLSA into no-reference image quality assessment and proposes a new no-reference image quality assessment algorithm based on PLSA.The algorithm integrates local bottom features of images in the first by producing reasonable visual-words.Then, it use statistical methods to establish PLSA model,the conditional probability distribution for "image-latent semantic-visual word", in order to find the latent semantic structure in the images and stride across the semantic gap.Finally, it acquire the result of the image quality assessment relying on the distribution of latent semantic of the images. Starting with the research on no-reference image quality assessment algorithm,this article proposes the image quality assessment algorithm based on PLSA for blurred image and the image quality assessment algorithm based on PLSA for Gauss noise image in turn.Both of the algorithms all achieved good results on the image quality assessment.Thus, this article then proposes a image quality assessment algorithm based on the PLSA for various distortion image.Compared with the former algorithm, this improves more better performance.