POSTECH Computer Graphics Laboratory

Render the future | Beyond the reality | Visual intelligence

About us Get Started

About Us

Our Research

The research goal of Computer Graphics Lab is to develop software and tools for various graphics applications. To achieve the goal, we are currently focusing on 3D reconstruction, face reconstruction, scene understanding, image registration, image enhancement, video processing and new media processing. Recently, we have applied machine learning techniques, especially deep learning, to many computer graphics applications. We always try to pursue a good quality of techniques and create new applications with an innovation and a creativity.

Research

3D Scene Reconstruction

RGB-D images, Texture optimization,
Non-rigid 3D reconstruction, Face reconstruction, …

Image and Video Processing

Deblurring, Upright adjustment, HDR, Texture filtering, …

Computational Photography with Big Visual Data

Super-resolution, Color enhancement,
Composition enhancement, Video stabilization, Personalization, …

New Media Applications

360 image and video processing, Augmented reality, …

Publications

See all publications

Texture Map Generation for 3D Reconstructed Scenes

We present a novel method for generating texture maps for large-scale 3D geometric models reconstructed using consumer RGB-D sensors. Photometrically consistent and sharp texture map can be used for various CG/VR applications.

Scale-aware Structure-Preserving Texture Filtering

With conventional edge-aware filters, it is often challenging to handle images of high complexity where features of multiple scales coexist. Our method keeps sharp edges and corners during the texture filtering using an adaptive per-pixel kernel scales.

Fast non-blind deconvolution via regularized residual networks with long/short skip-connections

We propose a novel framework for non-blind deconvolution using deep convolutional network. Given a blur kernel, our method restores a latent sharp image from a blurry image. Our method achieves better performance than the state-of-the-art methods with faster speed.

Robust Upright Adjustment of 360 Spherical Panoramas

In a spherical panorama, alignment of the scene orientation to the image axes is important, so we suggest an automatic method for upright adjustment of 360 spherical panorama images.