No abstract

Image inpainting techniques have been widely investigated to remove undesired objects in an image. Conventionally, missing parts in an image are completed by optimizing the objective function using pattern similarity. However, unnatural textures are easily generated due to two factors: (1) available samples in the image are quite limited, and (2) pattern similarity is one of the required conditions but is not sufficient for reproducing natural textures. In this paper, we propose a new energy function based on the pattern similarity considering brightness changes of sample textures (for (1)) and introducing spatial locality as an additional constraint (for (2)). The effectiveness of the proposed method is successfully demonstrated by qualitative and quantitative evaluation. Furthermore, the evaluation methods used in much inpainting research are discussed.

Abstract

Diminished reality aims to remove real objects from video images and fill in the missing regions with plausible background textures in real time. Most conventional methods based on image inpainting achieve diminished reality by assuming that the background around a target object is almost planar. This paper proposes a new diminished reality method that considers background geometries with less constraints than the conventional ones. In this study, we approximate the background geometry by combining local planes, and improve the quality of image inpainting by correcting the perspective distortion of texture and limiting the search area for finding similar textures as exemplars. The temporal coherence of texture is preserved using the geometries and camera pose estimated by visual-simultaneous localization and mapping (SLAM). The mask region that includes a target object is robustly set in each frame by projecting a 3D region, rather than tracking the object in 2D image space. The effectiveness of the proposed method is successfully demonstrated using several experimental environments.

Abstract.Recently, many types of omni-directional cameras have been developed and attracted much attention in a number of different fields. Especially, the multi-camera type of omni-directional camera has advantages of high-resolution and almost uniform resolution for any direction of view. In this paper, an extrinsic camera parameter recovery method for a moving omni-directional multi-camera system (OMS) is proposed. First, we discuss a perspective n-point (PnP) problem for an OMS, and then describe a practical method for estimating extrinsic camera parameters from multiple image sequences obtained by an OMS. The proposed method is based on using the shape-from-motion and the PnP techniques.

This paper proposes a new diminished reality technique which removes AR markers from a user's view image. In order to achieve natural marker hiding, three factors should be considered; (1) naturalness of texture generated on a marker area. (2) geometric consistency between consecutive frames, (3) photometric consistency between a marker area and its surrounding. In this study, assuming that an area around a marker is locally planar, the marker area in the first frame image is inpainted using the rectified image to achieve high-quality inpainting. The unique inpainted texture is overlaid on the marker region in subsequent frames according to camera pose for temporal geometric consistency. Global and local luminance changes around the marker are reflected to the inpainted texture for photometric consistency.

No abstract

Abstract. Estimating camera position and posture can be applied to the fields of augmented reality and robot navigation. In these fields, to obtain absolute position and posture of the camera, sensor-based methods using GPS and magnetic sensors and vision-based methods using input images from the camera have been investigated. However, sensor-based methods are difficult to synchronize the camera and sensors accurately, and usable environments are limited according to selection of sensors. On the other hand, vision-based methods need to allocate many artificial markers otherwise an estimation error will accumulate. Thus, it is difficult to use such methods in large and natural environments. This paper proposes a vision-based camera position and posture estimation method for large environments, which does not require sensors and artificial markers by detecting natural feature points from image sequences taken beforehand and using them as landmarks.