PhotoTechEDU Day 31 - Color Balance: Babies, Rugs & Sunsets
Google Tech Talks
December, 12 2007
Achieving pleasing color balance is one of the most important and
difficult problems in photographic systems; if the color balance is
off, other image quality attributes drop in important and the result
is unacceptable. In this talk I will discuss the differences between
color balance and white balance for both photographic and machine
vision applications and I will outline the literature of the subject.
I will explain some of the more basic and some of the more advanced
methods and relate these to complexity and system calibration issues.
I will give several examples that show how some methods can fail and
why some images can be extremely difficult. I will touch on the
relationship between color balance and color perception and how this
differentiates photographic systems from machine vision systems.
The goal of PhotoTechEDU is to have a Photographic Technology short course at Google. The course will teach Google engineers:
useful properties of light and image formation
theory and techniques of photographic optics and image capture
theory of colorimetry and techniques of color reproduction
where and how photography is being used in Google products and projects
what tools exist inside Google for photographic image storage, processing, etc.
and lots more...
The course will be serialized, meeting once a week for one hour, for at least 12 weeks, with probable extension as topics of interest are identified. Homework problems and projects will be suggested at various meetings, but they are completely optional. No formal class signup is necessary. You can find details on this wiki page: http://wiki.corp.google.com/tw...
Speaker: Paul Hubel
Dr. Paul M. Hubel has been working as Chief Image Scientist at Foveon, Inc. since 2002. His work includes the design of image processing algorithms and sensor architectures for high, middle, and low end cameras. Before joining Foveon, Dr. Hubel worked for ten years as a Principal Project Scientist at Hewlett-Packard Laboratories working on color algorithms for digital cameras, photofinishing, scanners, copiers, and printers.
Dr. Hubel received his B.Sc. in Optics from The University of Rochester in 1986, and his D.Phil. from Oxford University in 1990. His D.Phil thesis is titled "Colour Reflection Holography". As a graduate student, Dr. Hubel worked part-time at the Rowland Institute for Science under Dr. E.H. Land and later as a Post Doctoral Fellow at the MIT-Media Laboratory. Dr. Hubel has published over 30 technical papers, book chapters, and authored 25 patents. Dr. Hubel is a member of IS&T and SPIE, and has served as the technical and general chair of the IS&T/SID Color Imaging Conference.
Channel: People & Blogs
Uploaded: December 14, 2007 at 10:25 am
Author: googletechtalks
Length: 59:12
Rating: 4.60
Views: 3109
Tags: education engedu google googletechtalks talk talks techtalk techtalks
Video Comments
|
ler0np0 (March 17, 2008 at 4:14 pm)
I think you techies don't understand color manipulation manually by hands. RGB are for novices who mostly rely on one click automatic overall corrections by computers but the finest images are the ones custom adjusted by artists using the right and exact blend of CMYK.
albedoshader (March 12, 2008 at 10:38 am)
CMYK is like the worst color space for image manipulation.
ler0np0 (March 11, 2008 at 1:01 am)
I don't have major problems with color balance. Everything can be solved simply by converting the image to CMYK and tweaking the color channel curves using Photoshop.
ler0np0 (March 11, 2008 at 1:00 am)
I don't have major problems with color balance. Everything can be solved simply by converting the image to CMYK and tweaking the color channel curves using Photoshop.
albedoshader (January 26, 2008 at 11:44 am)
Bot at all! The color rendering pipeline is exactly what's taking place when maniplating images in photoshop. Things like Von Kries transformation matrices and cromaticity calculations are made and used by eg the adobe rendering engine when using appropriate color spaces. The end user simply doesn't see the maths behind it.
silverchevy (January 8, 2008 at 11:18 pm)
Waaay too technical for real world photography! I suppose if I wanted to build a digital camera, this would be required viewing. |
|