The Research On Color Cast Detection And Image Quality Assessment In Digital Image

With the fast developing of information technology and the Internet in recent years,a variety of imaging devices have been widely used.The digital image plays a more and more important role in human life,and the quality of digital image has been concerned by ordinary consumers as well as researchers.The color feature is a key factor to assess digital image quality.Therefore,the color cast detection algorithm in digital image is depth studied in this paper.On this basis,the comprehensive assessment method and verification testbed for image quality are designed and implemented.At first,the basic principle and realization process of the classical color cast detection methods are analyzed and studied in detail,this kind of method called color constancy algorithm,including White-patch algorithm,Gray-world algorithm,etc.Considering the limitations of color constancy algorithm,the color cast detection algorithm based on Lab chromaticity histogram is depth analyzed.Through the experiments and analysis,it is found that the definition of the color cast features is insufficient,which lead to the detection accuracy of the algorithm based on Lab chromaticity histogram is still low.Thus,the color cast detection method based on Lab three-dimensional color space is presented,which aiming at the shortcomings of the original algorithm based on Lab chromaticity histogram.Firstly,for the two-dimensional chromaticity histogram,a new color cast feature is proposed which the distribution characteristic in the height direction of chromaticity histogram is considered.What’s more,we further considering the feature change rule between the original image and the NNO image,then the feature change ratio is proposed to quantitatively characterize the change rule.Secondly,for the L component histogram,quantitatively characterize of regional aggregation distribution characteristics of L component by fitting the envelope curve of L component histogram,thus the color cast features in L channel are proposed.Finally,according to the EC features and the features proposed,the complete procedure of the color cast detection method based on Lab three-dimensional space is presented.The experimental results show that the color cast detection method proposed is effective,and it has good supplement and improvement for classic color cast detection algorithm based on Lab chromaticity histogram.What’s more,the thresholds can be adjusted easily when classify the intrinsic cast image and the true cast image,so as to adapt to more scene.In addition,an image quality assessment testbed is designed,including assessment metric and architecture,image quality multi index fusion assessment model,as well as hardware design and system software design.The testbed can be used for road traffic Closed-Circuit Television surveillance system,and the image quality of the whole CCTV surveillance system can be assessment in the testbed.