Study On Quality Assessment Algorithms Of The Aerial Image Major:Signal And Information Processing

The UAV has the characteristic of simple operation,low cost,and can reach the area where people cannot go to,and the images that used this method also has the advantage of high definition,rich in content and wider angle of view.So this technique had been widely used in environment,hydrology,geology,meteorology and so on.Aerial images of this paper are obtained by the UAV airborne camera equipment that shooting the inte:rested target which on the ground.In this test,the UAV in the altitude around 500 meters.The aerial images are mainly used in geological environment monitoring.In reality,the limits of the airborne hardware devices lead to us make a quality assessment for the image sequence before transferred them to the ground.Therefore,it has an important significance that we get one quality assessment method for the aerial images.This article deeply study and discuss the existing image quality assessment algorithms,and analysis the causes of image distortion,combining with image fuzzy principle,this paper present a method that use the thought of edge point spread to measure the level of image distortion.The main research contents are as follows:1.In this paper,a brief expound of the subjective assessment and the objective assessment of image quality,then analyzed the researches of the domestic and status quo and the direction of development,the no-reference image quality measuremernt is stressed.2.First the main factors that affecting the quality of aerial image are discussed.Then summari.zed and compared the existing image quality assessment algorithms for the no-reference.Finally,by studying the characteristics of the image-rich in detail,scene complexity,a larger picture size,and the requirement of the algorithm in the reality,we extracted the algorithm used in the paper.3.In order to improve the real-time performance of the algorithm,we blocked the image before edge detection,and according to human visual attention in different regions of the image exacting the region of interested.Secondly,we detected the interested area using the 8-direction sobel operator.During the edge detection,by analyzing the absolute value of 8-direction gradient for each edge point,we get the pixel value and direction of edge point.Finally,we processed those edge points in order to get less binary edge image of noise point.4.In the gray image,we used the determinate edge points and the direction of edge points to calculate edge width along the gradient direction.Considering the attention different to different images area,in this paper,the final result using different weights for different areas.The smaller of the result,the sharper and better of the image.In order to verify the accuracy of the algorithm,finally,using this algorithm,the average gradient algorithm,the Laplace algorithm and the neighbors domain algorithms to evaluated a sequence of images,we get their sort algorithm evaluation value.By compared the algorithm with the subject assessment,we have found that the algorithm is better than other algorithms on accuracy and timeliness.However,the assessment results are not very satisfactory for the distortion caused by heavy fog,and this is one of the places in need of improvement for further study.