Visual Perception Characteristics And Image Quality Assessment Via EEG

The subjective assessment of visual information is viewed through eyes to make an evaluation of the visual quality.However,due to the effect of various subjective factors during people's perception of images,the quality scores obtained by traditional subjective methods have great individual differences.Therefore,aiming at the different perception responses caused by images with or without significant targets,the image quality perception thresholds under different distortion levels,and perceptual changes in video quality caused by moving targets at different speeds,this thesis directly measures the EEG signals of subjects viewing different quality images or videos,to analyze the relationship between changes in EEG signals and visual perceptual characteristics.The main research results are as follows:(1)A significant target image perception quality assessment method based on EEG is proposed.Through analysis of the differences in image quality perception EEG signals,experimental results can be obtained that when the degree of image distortion increases,the impact of target on visual perception quality becomes more significant.Firstly,a group of simple background images without significant target and a group of images with an obvious target are selected as the stimuli.Then,subjective experiments on image quality assessment were designed to obtain two sets of quality-sensed EEG signals.Finally,the subjects' EEG signals are processed,analyzed and compared.Results show that the peak of P300 becomes larger as the degree of distortion becomes larger.At the same time,comparing the two groups of images,in the case of large degrees of distortion,the EEG signals triggered by the images with a significant target are significantly different from those without targets.(2)An image distortion perception threshold and quality assessment method is proposed.Through analysis of the obtained EEG signals,experimental results can be obtained that the change of the just noticeable distortion has a linear correlation with the degree of distortion.Firstly,a checkerboard pattern is selected as the visual stimulus.Five distortion levels are uniformly constructed and ten distortion parameters for each distortion level are selected to construct the distorted image stimuli.Then,the threshold range where the subject can notice the distortion change is determined,and the corresponding EEG signals are obtained.Finally,the five groups of EEG signals are comprehensively analyzed.The results show that the larger the degree of distortion,the greater the threshold for subjects noticing the change of distortion.The relationship of threshold and the distortion degree presents a linear correlation.(3)A video quality assessment method for different target movement speeds based on EEG is proposed.Through analysis of the video quality-sensed EEG signals,experimental results can be obtained that the faster the target interested moves,the earlier the P300 appears.Firstly,a natural scene video containing a moving target is selected,and four sets of video stimuli with different playback speeds are designed.For each speed,corresponding distortion level is designed.Then,according to the subjective experimental criteria for video quality assessment,experiments are designed to obtain EEG signals.Finally,the four groups of EEG signals are processed and compared.The experimental results show that at the same speed,as the degree of distortion increases,the peak value of P300 caused by video stimulation also increases.The latency of P300 caused by different speeds of the same distortion level are obviously different.The faster the playback speed,the shorter the latency.In summary,through qualitative and quantitative analysis of the EEG,visual perception characteristics are explored and image perception quality is calculated.Not only can it reflect the subjective feelings,but also include objective measurement,which will provide a physiological basis for the design of image quality assessment algorithms.