Translating the Viewing Position in Single Equirectangular Panoramic Images Frode Eika Sandnes 1,2 and Yo-Ping Huang 3 1 Department of Computer Science, Faculty of Technology Art and Design, Oslo and Akershus University College of Applied Sciences, Oslo, Norway 2 Faculty of Technology, Westerdals School of Art, Communication and Technology, Oslo, Norway 3 Department of Electrical Engineering, National Taipei University of Technology, Taipei, Taiwan 10608 email: Frode-Eika.Sandnes@hioa.no, yphuang@ntut.edu.tw Abstract—Equirectangular panoramas are popular tools for achieving 360 degree immersed viewing experiences. A panorama captures a scene from one point and panoramic viewers allow the user to control the viewing direction, but the viewer is not allowed to move around. This study proposes a strategy for transforming equirectangular panoramic images with the effect of moving freely in three dimensions. The strategy assumes that the panorama is an enclosed space comprising a flat ground and flat vertical walls. A equirectuangular Hough transform is proposed for detecting the boundaries of the respective planes. The panoramic image is then decomposed into the respective planes, the viewing point is translated and a new panoramic image based on the new viewing position is composed. Preliminary proof of concept test shows that the strategy allows free translation within simple panoramic images. Keywords-equirectangular panorama, Hough transform, translation, edge detection, immersed interaction I. INTRODUCTION Equirectangular panoramas have become popular especially with services such as Google street view [1], but are also used in other domains such as museums [2]. Equirectangular panoramas are single images that capture a scene viewed from one point going in all directions. Full equirectangular panoramic images are twice as wide as they are high as they represent 360 degrees horizontally and 180 degrees vertically. Panoramic images are viewed using standard interactive panoramic viewing software [3, 4]. The user can usually control the viewing direction and field of view giving the user a convincing impression of being present in the actual location. Equirectangular panoramas are captured either by special cameras lenses or multiple images with the camera pointing in the different direction. Special software is used to transform single images taken by a special lens, or merge multiple images, into the equirectangular panoramic image [5, 6]. Although panoramic viewers gives the observer a realistic immersed third-person experience, they do not allow the observer to move around. For example, Google street view allows the viewer to move around, but the viewer is moving from one panorama to another where the panoramas are captured at regularly spaced positions along the streets. The transitions between the different panoramas are discrete and sudden. The objective of this work is to develop a method that allows a viewer to move around certain panoramas. This work assumes that the panorama contains man-made objects, which often contain flat faces. Example scenarios include moving around inside a room defined by one panoramic image, or moving from one panoramic image to another in other to create more smooth transitions between different panoramic images. In practice, the planes would usually represent walls, being it inside a building or outside between buildings. The method uses a Hough transform to attempt identifying flat vertical faces, or walls, in the panorama. Using the face information the panoramic image is decomposed into the respective faces, the viewing position is moved and finally a new panoramic image is reconstructed using the flat faces according to the new viewing position. II. BACKGROUND Several approaches with similar goals have been attempted in the literature. Common to these methods is that they rely on two or more panoramic images. Shi et al. [11, 12] proposed a method for changing the viewer’s position in a location by interpolating between different cubic panoramic views. Cubic panoramas comprise a set of six square images representing the panorama in all six direction like a cube or box. Shi et al.’s interpolation approach is based on ray-tracing techniques, which checks the color consistency of corresponding pixels in different panoramas. Shanat and Laganiere [13] solved the same problem of interpolating between different panoramic views, but did it in real time with a graphical processing unit. Their approach is based on computing the optical flow field between adjacent cubic panoramas, and then morphing is used to achieve the motion between the panoramas. Both color features and geometric features were incorporated into the interpolation. Zhang et al.’s [14, 15, 16] approach to interpolating between different panoramas is based on computing a complex mesh of triangles, when are subsequently used to resynthesize panoramic 2016 IEEE International Conference on Systems, Man, and Cybernetics • SMC 2016 | October 9-12, 2016 • Budapest, Hungary 978-1-5090-1897-0/16/$31.00 ©2016 IEEE SMC_2016 000389