Aerial Image Quality Assessment

With the rapid development of science and technology,such as artificial intelligence,mobile Internet,and cloud computing,unmanned aerial vehicles have penetrated into all aspects of the society,both in civilian and military,and play an irreplaceable role.Because the storage space of the unmanned aerial vehicles is limited,both offline do follow-up studies,or online analysis,all need to pick an image with high quality,the subjective evaluation is time-consuming with low efficiency,so we need to use the objective quality assessment algorithm to automatically select the required character image.In actual scenarios,because of the different shooting conditions,UAV may carry different imaging devices.Therefore,this paper designs an image quality assessment algorithm for aerial images and infrared images,so that it can accurately evaluate the image quality.The work and innovation of this article are as follows:The database is the basis for quality evaluation,objective quality evaluation algorithm needs to rely on subjective quality scores to verify its performance,to construct a natural aerial image quality evaluation database and an aerial image of infrared image quality evaluation database.Through the analysis of the scene content characteristics of aerial images,selected 1158 representative aerial natural scene as a reference image,and simulate the distortion of aerial images of the actual natural environment,expand the 17370 images of the reference image based on image distortion;through the analysis of noise characteristics of infrared image acquisition in the process of simulation,7 Typical noise adding reference infrared image selected,composed of 700 amplitude distortion of infrared image.In order to make the vision system closer to people's subjective quality scores,according to the International Telecommunication Union video right evaluation standard,design professional and subjective experiments,mathematical methods for the corresponding treatment,had reliable subjective scores for each image in the database.According to the type of aerial image database distortion,analyze the changes of the distorted image signal-to-noise ratio and gradient information of different types of noise pollution in different intensity,the full reference of the natural image quality evaluation method of similarity bias ratio and gradient of a combined peak signal to noise of the.The difference method using different noise samples and the subjective scores and PSNR gradient amplitude change trend in the training process,the analysis of signal-to-noise ratio and gradient information in the measurement of different pollution image quality attenuation degree of accuracy,and to adjust the joint model parameters,then according to the model of aerial image quality.This method has low computational complexity,high accuracy,can guarantee the quality of aerial image accurately measure the real-time situation.Based on the characteristics of infrared image,a no-reference infrared image quality evaluation method based on the statistical characteristics is proposed.This paper in the spatial domain using NSS theory to capture the global features of the image,at the same time using the DCT transform of the image local information concerned,local characteristics of infrared image extraction in the DCT domain,then the complementary combination of NSS global statistical features and DCT domain local characteristics,the degree of attenuation of infrared image quality of a full range of accurate.The algorithm can not only evaluate the single noise but also evaluate the quality of mixed noise,and it has high accuracy and can meet the needs of industry.