Research On The Technology Of Panorama Visual Distortion Correction

Panorama is a large real image which is stitched by computer from multiple images captured by a camera from multiple angles. The building of panorama is a method of scene modeling based on image drawing, compared with the method of scene modeling based on geometry models. The former is faster, more realistic and needs less data, therefore is widely used in many areas, such as operations of arms, training simulation, showing of tours, introduction of houses and e-commerce and so on. Among them, the most prominent is the street view in conjunction with GIS technology, which has become one of the geo-information services that has far-reaching impact on people’s lives.At present, the visual perception problems existed in the panorama due to deformation is still not completely resolved. Panoramas still does not provide the users with accurate spatial cognition. This paper focused on the study of causes, characteristics and correct methods in visual deformation of a panorama. The major work includes the following aspects:1. This paper reviewed the course of development and major achievements of Panoramic Map, described the panorama visual distortion phenomenon, and analyzed the characteristics of panorama visual cognition. It put forward the needs of panorama visual distortion correction research, summarized the domestic and foreign status of the method of panorama visual distortion correction and panorama dynamic interactive technology development, which laid the theoretical foundation for the further research.2. This paper presented a kind of correct the panoramic visual deformation algorithm based on moving least square. It regarded any panorama projection as the spherical model to describe. During the processing of spherical projection to display graphic, it constructed the angle, line, area of visual distortion measure factor by using moving least squares according to the panorama vision cognitive characteristic. It minimized the visual distortion error, and controlled the deformation constraint of panoramic vision. It made sure the scene feature similarity, and corrected the linear bending in the scene.3. Based on panorama pyramid model the paper proposed single view panorama roaming method adapt to human cognitive characteristics. Corrected vision deformation of top panorama in Pyramid model, considering the changes caused by the modified Pyramid model, changed the viewpoint space from spherical to plane smoothly and designed view matrix between different roaming. It overcame the visual distortion caused by perspective projection during roaming.4. The TIP algorithm is proposed to adapt to the visual navigation method based on multi viewpoint panorama. It used the rendering to texture technology to construct TIP image space real-time, which overcome gap between the TIP image space and panoramic space. To achieve control of the roaming, the paper gave the calculation method of relative depth of TIP image space. It designed the stretch algorithm of TIP image texture by GLSL programming. Multi view roaming method based on TIP, in line with the forward vision, overcame the shortcoming of low resolution TIP facade, created the image space in the roaming process depth.