Research On Blind Camera Image Quality Assessment Based On Visual Perceptual Representation

With the development of electronic commerce and social networking tools,digital image is an important way for people to obtain information because it can express things intuitively.Because the image quality evaluation can be used to guide and optimize the image compression,image restoration,monitor image acquisition,.Therefore,the image quality evaluation is becoming more and more important both in the field of scientific research and in the practical application of image processing system.Since the full reference and reduced reference image quality assessment models need the whole or part of reference images,but the original images can not be obtained.in many practical applications.Hence no reference image quality evaluation algorithm has been widely studied.This paper will study the no-reference quality evaluation algorithm.This paper summarizes the research status of image quality assessment,and analyzes the related research of full reference quality assessment,reduced reference quality evaluation and no reference quality evaluation.Then it tries to make research on the quality assessment of blur images,noise images,multiply distorted images and camera images,and proposes four efficient algorithms for these objects.The main work and contributions of this paper are as follows:1.We develop a novel training free no-reference(NR)quality metric(QM)for noise images and blur images based on a unified brain theory,namely,free energy principle.The free energy principle tells that there always exists a difference between an input true visual signal and its processed one by human brain.The difference encompasses the “surprising” information between the real and processed signals.This difference has been found to be highly related to visual quality and saliency.More specifically,given a distorted image signal,we first compute the aforesaid difference to approximate its visual quality and saliency via a semi-parametric method that is constructed by combining bilateral filter and auto-regression model.Afterwards,the computed visual saliency and a new natural scene statistic(NSS)model are used for modification to infer the final visual quality score.Extensive experiments are conducted on popular natural scene image databases and a recently released screen content image database for performance comparison.Results have proved the effectiveness of the proposed blind quality measure compared with state-of-the-art no-reference QMs.2.Blur plays an important role in the perception of camera image quality.Generally,blur leads to attenuation of high frequency information and accordingly changes the image energy.Quaternion describes the color information as a whole.Recent researches in quaternion singular value decomposition show that the singular values and singular vectors of the quaternion can capture the distortion of color images,and thus we reasonably suppose that singular values can be utilized to evaluate the sharpness of camera images.Motivated by this,a novel training-free blind quality assessment method considering the integral color information and singular values of the distorted image is proposed to evaluate the sharpness of camera images.The blurred camera image is first converted to LAB color space and divided into b locks.Then pure quaternion is utilized to represent pixels of the blurred camera image and the energy of every block are obtained.Inspired by the human visual system appears to assess image sharpness based on the sharpest region of the image,the local sharpness normalized energy is defined as the sharpness score of the blurred camera image.Experimental results have demonstrated the effectiveness of the proposed metric compared with popular sharpness image quality metrics.3.The current image quality metrics work on the assumption that an image contains single and simulated distortions which are not representative of real camera images.In this paper we address the problem of quality assessment of camera images from two respects,natural scene statistics(NSS)and local sharpness,and associated three types of features.The first type of four visual perceptual representation features measures the naturalness of an image,inspired by a recent finding that there exists high correlation between structural degradation information and free energy entropy on natural scene images and this regulation will be gradually devastated as more distortions are introduced.The second type of four visual perceptual representation features originates from an observation concerning the NSS that a broad spectrum of statistics of distorted images can be caught by the generalized Gaussian distribution(GGD).Both the two types of features above belong to the NSS based models,but they come from the considerations of local auto-regression(AR)and global histogram,respectively.The third type of three visual perceptual representation features focuses on estimating the local sharpness,which works by computing log-energies in discrete wavelet transform domain.Finally our quality metric is achieved via a SVR-based machine learning tool,and its performance is proved to be statistically better than state of-the-art competitors on the CID database dedicated to the quality assessment of camera images.4.As an extension of Discrete and Complex Wavelet Transform,Quaternion Wavelet Transform(QWT)has attracted extensive attention in the past few years,because it can provide better analytic representation for 2D images.The QWT of an image consists of four parts,i.e.,one magnitude part and three phase parts.The magnitude is nearly shift-invariant,which characterizes features at any spatial location,and the three phases represent the structure of these features.This indicates that QWT is more powerful in representing image structures,and thus is suitable for image quality evaluation.In this paper,an efficient and effective Camera Image Quality Metric(CIQM)is proposed based on QWT,which is utilized to describe the intrinsic structures of an image.For an image,it is first decomposed by QWT with three scales.Then,for each scale,the magnitude and entropy of the subband coefficients,and natural scene statistics of the third phase are calculated.The magnitude is utilized to describe the generalized spectral behavior,and the entropy is used to encode the generalized information of distortions.Since the third phase of QWT is considered to be texture feature,the natural scene statistics of the third phase of QWT is used to measure structure degradations in the proposed method.All these visual perceptual representation features reflect the self-similarity and independency of image content,which can effectively reflect image distortions.Finally,random forest is utilized to build the quality model.Experiments conducted on three camera image databases and two multiply distorted image databases have proved that CIQM outperforms the relevant state-of-the-art models for both authentically distorted images and multiply distorted images.