Yong-In Yun

A new shape descriptor using sliced image histogram for 3D model retrieval

3-dimensional (3D) model retrieval is one of emerging research fields to find the matching shape of a given query from 3D database. This paper proposes a new shape descriptor for 3D model retrieval. A shape descriptor of 3D model requires a model normalization in order to be invariant to translation, rotation, and scale for its model. Our method is based on principal component analysis (PCA) for normalizing all the models. The shape descriptor is using a histogram of 2D images sliced along the x-, y-, and z-coordinates for measuring the similarity in 3D models. Sliced shapes for a 3D model involve a hundred planes to orthogonalize with the x-y-, and z-coordinates, respectively. Therefore, sliced shape is the 2D plane images created by intersecting at between 3D model and the planes. Our approach is to compute the slices of 3D model for the x-y-, and z-coordinates, respectively and to set by the principal axis based on principal component analysis (PCA) in order to match the 3D model between the given query and the database. Experimental results show that the proposed approach outperforms the previous approaches. We demonstrate the proposed 3D retrieval system with an intermediate example at each step.

Detection of digital forgeries using an image interpolation from digital images

Most of digital forgeries are created by Adobe Photoshopreg. Detecting digital forgery is one of most important emerging technology to solve digital forgery of digital images taken by a digital camera. This paper proposes a new method for detecting digital forgery using an image interpolation technique from digital images. When digital forgery is produced from digital images, it always includes the interpolated image between the images. Our approach is to compute the interpolated coefficient for the images using expectation-maximization (EM) algorithm. We demonstrate detecting the forgery images used to each filter for the tool of Adobe Photoshopreg and show a ratio of detecting the forgery images.

A new estimation approach of resampling factors using threshold-based peak detection

This paper proposes a new method for digital image forensics using threshold-based peak detection to estimate interpolation factor. We describe how resampling introduces periodic correlations, and to detect these correlation using the differential method along the horizontal, vertical and diagonal direction, respectively.

Three-dimensional reconstruction of an object using three-planar homography of a single image

Camera calibration is essential in order to perform 3-D reconstruction from uncalibrated images. This paper proposes a new technique of camera calibration using a homography among three planar patterns taken from a single image. It is implemented more easily and simply than conventional methods. Experimental results support the hypothesis and show that the performance of the proposed method is better than that of conventional methods. Examples of 3-D reconstruction are demonstrated using the proposed algorithm on sequences of images.

3D pose estimation for foot motion tracking from image sequences

This paper proposes a new method for foot pose estimation using a node of shoes and the planar ground in order to estimate diverse foot shapes from image sequences. We demonstrate foot motion tracking a scenario using the proposed algorithm for pose tracking of the human foot.

A new interface using image-based foot tracking for motion sensing devices

Foot interfacing has applications in games that require foot manipulation control using motion sensing devices. However, when part of a foot overlaps itself (selfocclusion), is out of the line of sight of the infrared sensor and it is then difficult to accurately capture foot pose. To overcome this problem, we propose a real-time motion sensing method based on 3-dimensional motion estimation. Our method employs pose parameters that are acquired using the ground plane and camera coordinates. This enables an assessment of the positions of the two joints that comprise a skeletal foot model. The proposed approach can capture a foots 3-dimensional information that an infrared sensor may miss; therefore, motion sensing devices pose estimation results improve. According to the tracking and experimental results, the proposed method is applicable for use by motion sensing devices.

Automatic Control Signal Generation of a CCD Camera for Object Tracking

CCD cameras monitoring a moving object generally operate under fixed view point or a defined pattern. So a free moving object goes out of the CCD camera’s sight in a short time and automatic control is required to observe it continuously. This paper proposes a signal generation algorithm for the automatic control of the CCD camera. The proposed algorithm computes horizontal and vertical displacements between the detected object position and the image center and coverts them into angles. Finally pan/tilt data are generated from the angles. To estimate the proposed algorithm, we compare the generated data in the automatic control with the measured data in manual control and carry out the object tracking using a simple object. The experiment results show that the difference between the generated and measured data is a negligible quantity and the object, which is moving in ±52 o /±40 o of pan/tilt, is kept in about ±13 o /±10 o .

Foot motion tracking for motion capture from stereo cameras

This paper proposes a new motion capture strategy to track foot motion from stereo cameras. We demonstrate motion capture of 3D foot model in real-time and show the performances of the proposed algorithm for the rotation and the translation of 3D foot models.

Full three-dimensional reconstruction using key-frame selection under circular motion

Fundamental-matrix and key-frame selection constitute one of the most important techniques for full 3-D reconstruction of objects from turntable sequences. This paper proposes the new methods for these selection problems in 3-D reconstruction from uncalibrated sequences taken with a turntable and the Fotonovo camera system. Also, we propose a projection-matrix refinement for accurate full 3-D reconstruction. Our approach utilizes single-axis motion. To evaluate the fundamental matrix, camera calibration and 3-D registration are generally needed. Our main contribution is a method for robustly determining the corresponding points between two images, and for accurately filling gaps in a sparse object so as to make surface reconstruction tractable. We do not need all frames, but only few pairs of images (key frames). The key-frame selection has the advantage in camera pose estimation and 3-D scene reconstruction of reducing the computational costs. Experimental results on real image sequences demonstrate accurate object reconstructions and robustness of the proposed methods.

A novel feature descriptor for image retrieval

This paper proposes a new color feature descriptor using double color histogram and color changing ratio for image retrieval. Specifically, a pre-processing is conducted on a given image for reducing the noise and data size dealt with. Next, the given color image is divided into fragment and particle regions according to color coherency. Then, color histogram and color changing ratio are used for computing a similarity measure in this paper. Experimental results show that the proposed approach outperforms the other conventional ones with respect to processing speed and retrieval accuracy.