The Application Of The Big Data Of Eye Tracking In The Optimization Of Image Quality

Human’s visual system is a kind of multi-target tracking system. To offer the brain easy access to key information in the minimum time, human’s eyes will focus on only a handful of areas or objects, the behavior of which is often called visual attention. And these areas or objects in the scene make up the Region of Interest(ROI). Extracting the ROI is the key research orientation of various disciplines such as image processing, image enhancement and machine learning. It can greatly reduce the time of image compression and coding, target recognition and image matching, therefore effectively improving the efficiency of information processing.The subjects of getting ROI in laboratory are limited, the amount of experimental data is small and the equipment is too expensive, all of which lead to the result that the obtainment of data cannot spread to all walks of life. Therefore, this paper discusses how to take advantage of public resources to collect the big data of eye movement, then introduces the data capturing process, and finally puts forward a way to use the big data of eye movement in improving the visual quality of image. The main work is as follows.Initially, in order to widely collect the data of eye movement and to increase the number of subjects, the paper improves virtual reality goggles which is modeled on eye movement instrument. On the premise of not affecting the users, the device can collect the image of eye movement by using infrared light source, and use the viewpoint tracking technology to get line of sight. The obtained viewpoint can be mapped to the position of real images through the transformation of coordinates.Moreover, after obtaining the big data of eye movement, this paper analyzes them using improved k-means algorithm and two-dimensional Gaussian simulation to get the real ROI. In consideration of the shortage of existing k-means algorithm, this paper makes some improvement and successfully identifies ROI using the improved k-means algorithm. At the same time, the paper points out that what the k-means obtains is the round area of ROI. Thus, by using two-dimensional Gaussian simulation, the mountain height map can be built, and then the irregular ROI area can be gotten by using the method of "cutting the top of the mountain".Finally, according to the feature that the central sunken area of human’s vision is the most clear, this paper puts forward the method of using the big data of public resources to obtain the data of eye movement, analyzing the data to get ROI, and then optimizing the visual quality of images. Based on this, the image quality can be adjusted dynamically to match different bandwidth. Therefore, the space of downloads can be saved and the problem that users have to make a comprise between the bandwidth and the image quality can be solved.This paper, based on previous studies, combined with the advantages of the era of big data, creatively proposes the use of public resources in optimizing the quality of image and the use of optimized images in attracting public resources, which is a positive circle. This paper expands the vision of research field and points out the future research direction.