Research On Quality Evaluation Technology Of Mosaic Images

With the development and maturity of image mosaic technology,evaluating the quality of mosaic images objectively to compare the performance of image mosaic algorithms becomes more and more important.At present,the image quality evaluation method for mosaic images is rarely studied,and the quality evaluation method for mosaic images is usually limited by standard reference image or mosaic algorithm.In the actual evaluation for mosaic images,it is of practical significance to study a general and independent quality evaluation method for mosaic images in face of complex mosaic scenes.Based on this,this paper focuses on the actual demand,and mainly studies the application of image quality evaluation technology in one-dimensional and two-dimensional mosaic images.By constructing mosaic image data set,the characteristics of mosaic images are analyzed.Then,combined with the principle of image mosaic and the law of visual perception,this paper designs and validates evaluation methods for sequence,distortion,alignment and fusion of mosaic images,and summarizes the calculation formula of quality evaluation value for mosaic images.Finally,a general and independent quality evaluation method for mosaic images is realized.The main research work of this paper is as follows:1.Sequence evaluation is the basis for the quality evaluation of mosaic images.Referring to the process of human visual evaluation,this paper proposes a sequence evaluation method based on feature points.By using SURF feature point extraction and KNN feature point matching algorithms,the ratio between feature points and matching points of the original small image to be stitched and the cropped region of the mosaic image to be evaluated is compared,and then the sequence evaluation value is calculated.2.Distortion evaluation is a necessary step for the quality evaluation of mosaic images.Combined with the principle of image mosaic,this paper proposes a distortion evaluation method based on the homography matrix.By using the RANSAC algorithm,the homography matrix is solved.The correspondence between matrix parameters and distortion degree is analyzed through experiments,and then the distortion evaluation value is calculated.3.Alignment evaluation is the key step for the quality evaluation of mosaic images.This paper proposes a alignment evaluation method based on similarity.According to the homography matrix,the mapping transformation image is estimated,and the alignment degree of the overlapped areas between the mapping transformation images is mainly studied.Then,by constructing the differential absolute value image,the alignment evaluation value is calculated.4.Fusion evaluation is the step to supplement and improve quality evaluation of mosaic images.This paper proposes a fusion evaluation method based on spatial frequency.The spatial frequencies of the mapping transformation image and the corresponding fusion area image are calculated.Then,by analyzing the corresponding relationship between the spatial frequency difference and fusion effect,the fusion evaluation value is calculated.A unified quality evaluation value of mosaic images is obtained by integrating the evaluation values of sequence,distortion,alignment and fusion.Combining with the subjective quality evaluation method,the quality evaluation method for mosaic images in this paper is used to carry out quality evaluation experiments of mosaic images in the data set,and finally the accuracy of the quality evaluation method for mosaic images is verified.In this paper,the project experiment is completed under the visual studio 2013 compiling environment.The experimental results show that the same results with the subjective evaluation can be obtained by using the quality evaluation method for mosaic images in this paper,which provides an innovative and feasible research idea.