This paper presents a framework that enables a user to manipulate his/her face shape three-dimensionally in video. Existing face manipulation applications and methods have some limitations: single photo, manipulation on image domain, or limited deformation. In the proposed framework, face is tracked from video by using landmark tracking and fitting 3d morphable face model to image, and the face model is further deformed according to the user input with mesh deformation method and rendered with texture from the frame image onto the camera preview. Therefore, unlike conventional applications and researches for face manipulation, the proposed framework allows the user to perform free-form 3d deformation of face in video and to view the deformed face at various viewpoints.

In conventional imaging, the geometry of image formation is described by the pinhole camera model, and it allows multiview reconstruction based on the perspective projection. If a light ray passes through different media, however, the pin-hole camera model becomes invalid due to refraction, causing changes in direction. In this paper, we present a method for 3D trajectory reconstruction under refraction at a cylindrical surface. In the proposed method, the refractive geometry is represented by a geometric mapping from image points to refracted rays, instead of physically modeling it. 3D points inside the cylindrical surface are estimated by computing intersection points between segments of corresponding refracted rays in multiview. In our experiments, we demonstrate that 3D trajectories of multiple moving objects inside a transparent cylinder, which is filled with water, are successfully recovered using multiple cameras.

This paper presents a method for interactive deformation of a real object. Our method uses a predefined 3D model of a target object for tracking and deformation. A camera pose relative to the target object is estimated using 3D model-based tracking. Object region in camera image is obtained by projecting the 3D model onto image plane with the estimated camera pose, and a texture map is extracted from the object region and mapped to the 3D model. Then a texture-mapped model is rendered based on a mesh deformed by user via Laplacian operation. Experimental results demonstrate that our method provides user interactions with 3D real objects on real scenes, not augmented virtual contents.

Since the minimum feature size of dynamic RAM has been down-scaled, several studies have been carried out to determine ways to protect cell data from leakage current in many areas. In the field of testing, more appropriate test algorithms are required to detect weak cells with leakage-current sources. In this paper, we propose an interleaving test algorithm that takes into account the equal bit-line stress regardless of the cell location. The proposed test algorithm allows screening of weak cells that cannot hold cell data due to the subthreshold leakage current. During the stress period, the algorithm can also detect other leakage currents. This paper presents the maximum stress differences according to the cell location, and determines the influence of the refresh operation on the maximum stress time. Therefore, this paper suggests a correlation between the refresh and read time to give maximum stress time.

This paper presents 3-D visual tracking on a mobile phone, which is one of the next promising challenges of visual tracking for mobile augmented reality applications. Our approach is based on model-based tracking using a 3-D model of an object on a target scene. A camera pose is estimated by iteratively updating infinitesimal motions of a camera, explicitly represented by 3-D rigid-body transformations. The 3-D visual tracking is implemented on recent Android smartphones. Its feasibility is shown by experimental results and discussed along with its performance. To our knowledge, this is the first work that demonstrates the 3-D model-based tracking on a mobile phone.

This paper proposes an augmented reality interface that provides natural hand-based interaction with virtual objects on mobile phones. Assume that one holds a mobile phone in a hand and sees the other hand through mobile phone's camera. Then, a virtual object is rendered on his/her palm and reacts to hand and finger movements. Since the proposed interface does not require any additional sensors or markers, one freely interacts with the virtual object anytime and anywhere. The proposed interface worked at 5 fps on a mobile phone (Galaxy S2 having a dual-core processor).