Image Quality Objective Assessment Research Based On Fourier Transform Domain

When Obtaining digital images and videos, images will inevitably be polluted bydifferent levels and types of distortions during image acquisition, compression,processing, transmission and replication because of a variety of physical and economicfactors like the optical properties of the apparatus for obtaining image, instrumental orenvironmental noise, processing technology, storage, bandwidth, etc. These distortionsdirectly affect the viewer’s visual perception of the image and the image informationextraction and understanding for people, that is to say the image quality has declined andit is difficult to meet actual needs of people. Therefore, it becomes necessary to provideevaluation methods of image quality for monitoring and evaluating various links thatimage degradation may occur.Image quality assessment is based on the viewer’s subjective feeling, so no matterwhat kinds of objective evaluation method was adopted, we should strive to make theobjective assessment results of various evaluation models in line with people’s subjectiveevaluation as much as possible, that is to say that the objective evaluation should take thesubjective evaluation as a criterion. Compared with subjective evaluation methods,objective evaluation methods have many advantages, like speed, low cost, wideapplication, unaffected by subjective factors, the evaluation results that can bereproduced etc. However, this kind of method still has drawbacks. So far, a variety ofproposed objective evaluation models have limited ability to mimic the human visualsystem, and it often appears that the results of objective evaluation is inconsistent withthat of subjective evaluation.On the basis of the relative theories and techniques of image quality evaluation andthe importance of phase in image processing, combined with structural similarity theory,this paper presents the discrete Fourier transform domain image based image qualityobjective evaluation model-PMSIM(Phase-Magnitude Similarity) and this modelprovides a solution for image quality evaluation.This paper presents two solutions of image quality objective evaluation models:PMSIM-1and PMSIM-2, and their performance were compared and analyzed. The twosolutions differ in the definition formulas of Phase-Magnitude Similarity. Phase-Magnitude Similarity of the first solution is defined by multiplying exponentialphase by exponential magnitude, yet,Phase-Magnitude Similarity of the second solutionis defined by the sum of the weighted phase and the weighted magnitude. For the secondsolution, the formula of the model under blocking conditions was also introduced. Bycomparing the experimental data, we can know that the model with the block size in thefirst solution has a better performance. In this case, this model was compared with otherclassic image quality evaluation models and it can be found that PMSIM model has abetter comprehensive performance.