Image Super-resolution And Quality Assessment Based On Visual Perception

In practical applications,the high resolution images are often difficult to obtain due to the limitation of hardware devices in the imaging system.To improve the resolution of the image by improving the hardware equipment is expensive and faced with some technical problems that are difficult to overcome in the short term of imaging system,so it is great significance that improve the image resolution by the software way.Super resolution image reconstruction technique is the use of signal processing technology,from single or multi frame low resolution image in reconstructed high resolution images of good quality.In this paper,we mainly study the single low resolution image reconstruction and its quality evaluation,the improvement to the existing method based,combined with the characteristics of human visual perception,consistent with human subjective preference for the high resolution image and image quality evaluation standard.The human visual system is a highly complex and intelligent information processing system,and can process image in a very short period of time.Combining the visual perception characteristics of human visual system with the computer image processing algorithm can effectively improve the processing efficiency of the latter.In the early stage of visual information perception,the human visual system is not equal to all image regions to be processed,but through the visual attention mechanism to select the area of interest in the first interpretation.The visual attention based image saliency detection algorithm can effectively reduce the content of the image to be processed and improve the efficiency of image processing.Due to the limited resolution of the visual system,the human eye cannot detect the signal content under a certain threshold.We can utilize this feature to eliminate the variationalinformation that has no effect on the human visual system,and enhance the consistency of objective image quality evaluation index and subjective evaluation.This paper focuses on the research of super resolution image reconstruction based on visual perception and the method of quality evaluation,the main contents and contributions include:1.The traditional interpolation algorithm is easy to cause the edge blur,but the human visual system is usually easy to notice the edge of the object in the image.And the traditional interpolation algorithms cannot deal with noisecommendably.As to the visual masking effect,noise in the flat area is more likely to attract people's attention.In view of the above problems,a super resolution reconstruction method based on edge focusing and adaptive filtering method is proposed.In this method,the original high resolution image is obtained by the traditional interpolation method,and then the image edge pixelis located by edge focusing measure.Adaptive filtering method is adopted to adjust the filter parameters automaticallyfor different pixels,in order to get the best filter result and achieve the final high resolution reconstruction image.Meanwhile,we propose a fast image block search algorithm,which is used to speed up the filtering method.Experiments show that this method can obtain better performance in both visual and objective evaluation.2.The visual attention mechanism of human visual system determines that people always give priority to the image contents of the region of interest when observe an image.But in the case of limited computing resources and high real time,improving the priority of the interesting region can speed up the algorithm of super resolution image reconstruction under the condition of ensuring the reconstruction quality of the interest area.So we propose a saliency model based super resolution image reconstruction method,this method has excellentadaptation and expandability.In addition,traditional interpolation method directly to the low resolution image pixels as the high resolution image in the corresponding position of the pixel processing,have nothinkabout the impact of reducing factor such as the fuzzy,down sampling and noise interference in the actual imaging process.To overcome this problem,we also propose a block interpolation method based on local structure similarity of images.3.As the problem of inconsistency between the image size and the low resolution image with the input of the super resolution reconstruction image,the existing widely used full reference image quality assessment standards are not applicable to the quality evaluation of super resolution reconstruction image and there is little research on the quality evaluation of super resolution reconstruction image.In view of the above problems,we proposed a super resolution image quality evaluation criterion based on the structure similarity and edge blur.On the one hand,the structure similarity reflects the fidelity of the reconstructed image to the reference image,on the other hand,from the edge of the fuzzy degree reflects the human visual system for the quality of the reconstructed image of the objective evaluation,combine of them to obtain the final evaluation index.The experiment results demonstrate that the proposed image quality evaluation criteria can be applied to the quality evaluation of super resolution reconstruction imagepreferably.The aforementioned research results improve the existing super resolution image reconstruction and quality evaluation method from the characteristics of human visual perception,and with forward-looking and challenging,have a certain theoretical significance and practical value.