Rong-Chi Chang

Photo defect detection for image inpainting

Image inpainting (or image completion) techniques use textural or structural information to repair or fill damaged portion of a picture. However, most techniques request a human to identify the portion to be inpainted. We developed a new mechanism which can automatically detect defect portions in a photo, including damages by color ink spray and scratch drawing. The mechanism is based on several filters and structural information of damages. Old photos from the authors family are used for testing. Preliminary results show that most damages can be automatically detected without human involvement. The mechanism is integrated with our inpainting algorithms to complete a fully automatic photo defects repairing system.

Motion picture inpainting on aged films

Video inpainting uses spatial-temporal information to repair defects such as spikes and lines on aged films. We propose a series of new algorithms based on adjustable thresholds to repair different varieties of aged films. The main contribution is an automatic spike and dirt detection mechanism. We prove that if appropriate threshold is once decided by the author, most damages in an aged video clip can be detected. In addition, the repairing procedure first estimates temporal information and obtain replacement blocks among several frames. Spatial information is then used to repair damages that can not be fixed by temporal information due to fast motion. The results are visually pleasant with most defects removed.

A Strategic Decomposition for Adaptive Image Transmission

Progressive image transmission (PIT) transmits the most significant portion of a picture, followed by its less important portions. The mechanism can be used in Web- based applications while users are browsing images. However, most PIT methods use the same pixel interpolation scheme for the entire picture, without considering the differ- ences among image blocks. This paper analyzes the efficiency of pixel interpolation schemes and test several decomposition mechanisms. The contribution results in an adaptive image transmission scheme, which takes the differences of picture portions into consideration. Moreover, this study tested 200 pictures in different categories and pa- rameters. In consequence, the overall bit rates can be reduced significantly with good PSNR values and user satisfaction. The visual result is superior to progressive JPEG on both objective (quantitative) and subjective (human) measures. An error recovery pro- cedure is also implemented in case that the transmitted pictures need to be fully recov- ered.

Multilayer Inpainting on Digitalized Artworks

Image inpainting automatically restores damaged images and partially removed pictures. Since most inpainting mechanisms inpaint damaged portions in a single layer, this study proposes a multilayer inpainting mechanism by examining how Chinese paintings are drawn in layers. The proposed multilayer inpainting mechanism employs a layer fusion strategy to detect the optimal inpaint combination among layers to restore paintings. Because this multilayer multi-resolution strategy considers damages in each layer from a multi-resolution perspective, it is superior to several existing techniques for restoring Chinese and Western paintings. In this study, the proposed algorithm is tested on more than 1,500 still images, with evaluations showing the effectiveness of image inpainting. The results indicate that the proposed algorithm achieves high PSNR values as well as high user satisfactions, including inpainting in some extreme cases where more than ninety percent of a painting are destroyed.

Spatial-Temporal Method for Image Restoration in Aged Motion Pictures

This paper proposes a series of defect detection algorithms for video in painting. The proposed algorithms are developed based on adjustable thresholds and the primary focus of these algorithms is to provide a better way to repair different kinds of aged films. Two major defect detection techniques that have been created in this study are automatic spike and dirt detection mechanism. Several important findings have been released in the study. First, the findings indicate that if the user is able to choose appropriate threshold, most damages in an aged video clip can be detected. Second, the findings exhibit that spatial information can be used to repair certain damages, which can not be fixed by temporal information due to fast motion. Finally, the results of this study are visually pleasant with most defects removed.

Automatic Web-Based EFP Studio System with XML Documents

The rising wave of internet applications has led to the rapid development of streaming media technology. This paper proposes time-triggered events algorithm using a script object transformation to XML documents to create the automatic EFP studio system for real-time editing and broadcasting of audio/ video on media server. The system provides the user access to web multiple cameras and AV equipment through the internet to execute EFP tasks. The system can perform tasks such as to web multipoint remote recording, switch cameras, separate takes, change scenes, remove background, edit, integrate, add special effects on videos, the functionalities of a digital switcher, while getting superb video quality, and so on. At the same time, it can conjugate with web site to construct personal network TV station, distance video learning and TV news for real-time broadcast.

A VR-based hyper interaction platform

We present a hyper-interaction platform which integrates several newly developed techniques, including motion classification, a motion reaction control mechanism and a high precision 3-D model of the International Space Station. The goal is to build an interaction platform for trainers to navigate in a virtual reality. The platform can be applied to customized mission training. The system integrated visualization environment, underlying 3-D model, and a human body tracking mechanism. The tracking system will be extended to 3-D coordination reconstruction and thus behaviors of users can be precisely identified. These navigation parameters are used in controlling navigation in the 3-D virtual environment. Another important issue of this system is to define a model for motion reaction. The definition of reaction can be applied to avatars or to the virtual environment. These integrated technologies can also be used in other virtual reality environments

Video inpainting and restoration techniques

Aged films may contain defects such as spikes or dirt, as well as long vertical defect lines. These defects were produced in file development or due to improper maintenance of films. We present a series of algorithms, which can detect and restore defects. In addition, the restoration technique is used with a motion tracking mechanism. Objects can be removed and holes can be inpainted. We aim to demonstrate three tools for 1) spike detection and restoration, 2) long vertical line detection and repairing, and 3) object removal and inpainting. Results of restored video clips show that our mechanisms are practical with good inpainted video quality.