Naked Eye Stereoscopic Image Playback Technology Based On Human Eye Tracking

Matching is the core step in binocular stereo vision.Stereoscopic vision is now developing hot.and it has developed in many fields.In this paper,a new algorithmic process is proposed to detect the depth information of the viewer,which has good effects in terms of accuracy and engineering time.Face recognition and human eye detection are important pillars of modern development,so it is necessary to study related technologies-This paper proposes a positioning technique based on Adaboost and ASM hybrids,and describes the process of weak classifiers cascading into strong classifiers and some key principles in ASM algorithm.Finally,based on the above principles,a set of stereoscopic display system is constructed,and the actual core principle is given,which achieves good results.The main features and contributions of this paper are as follows:(1)Propose a new stereo matching process.First,the left and right images are census transformed to obtain a disparity map.On this basis,the left and right consistency detection is performed to obtain the initial parallax,and the abnormal points and occlusion points appearing in the image are removed.In the algorithm,the left image is taken as an example to optimize the watershed algorithm to segment the image.In the segmented region,the abnormal point is further removed according to the distribution of the parallax.And using the method of plane fitting,the parallax value in the region and the coordinate axis of the pixel are modeled to obtain the equation of the plane.Then use the established plane equation to further calculate the invalid disparity value.In image segmentation,the image is first divided into relatively small regions,and the acquired parallax can be defined on the same plane.In the determined disparity plane,the culling of the abnormal point and the estimation of the invalid disparity are again performed as the training data of the next plane fitting until the initial disparity does not contain the invalid disparity value.(2)Introduction about implementation of algorithm for face recognition and human eye tracking.First,the image captured by the camera is detected by the skin color,in order to reduce the range of the search.Then use the Adaboost algorithm for face detection.Then on the basis of the above,the eyes are searched and tracked in the detected face area.This feature is implemented using the Opencv framework to reduce implementation complexity and improve accuracy.(3)Design a set of the naked eye 3D display playback system.Combined with the previous contribution points,a 3D display framework based on human eye tracking and stereo matching is designed.This chapter first introduces the specific principle of stereoscopic display,then describe the implementation process of the system,and finally give the actual effect.