HKUST Library Institutional Repository Banner

HKUST Institutional Repository >
Electronic and Computer Engineering  >
ECE Master Theses >

Please use this identifier to cite or link to this item:
Title: Automatic white balancing for digital camera and fast motion vector re-estimation for arbitrary downscaling of compressed video
Authors: Lam, Hong Kwai
Issue Date: 2004
Abstract: With the rapid growth of digital image technologies, there are many new features in digital cameras making them very popular these days. Among these features, AWB is one of the important features in digital cameras. The purpose of AWB is to compensate for the color shifting introduced by light source. If a digital camera does not have white balancing function, a captured white object will appear reddish when under a low color temperature light source and bluish when under a high color temperature light source. With AWB, a digital camera can automatically compensate for the color shift due to the different color temperatures of the light source. In this thesis, a novel AWB algorithm is proposed using information of luminance and standard deviation of RGB channels. From the experimental results in term of Euclidean distance of uniform perceptual color space, the proposed algorithm SDLWGW provides a good alternative solution to solve the AWB problem. A topic in fast compressed video downscaling is also presented in this thesis. Due to limited storage capacity and transmission bandwidth, most of the stored videos are in different compressed formats. In some applications, these pre-compressed videos need to be transcoded into different resolutions and bit-rates according to the display size of the target device and the available bandwidth. In consumer electronics market, there are different types of handheld device. These devices usually have different display sizes and also limited memory size. Therefore, when a pre-compressed video file is sent out from a video server to these devices, the resolution of the compressed video usually need to be changed. Since the straightforward brute force approach is computationally intensive, a fast algorithm is needed especially for real time application. A fast algorithm based on SVF is proposed to provide a good solution to solve this problem.
Description: Thesis (M.Phil.)--Hong Kong University of Science and Technology, 2004
xiv, 60 leaves : ill. (some col.) ; 30 cm
HKUST Call Number: Thesis ELEC 2004 LamH
Appears in Collections:ECE Master Theses

Files in This Item:

File Description SizeFormat

All items in this Repository are protected by copyright, with all rights reserved.