[Home]

Introduction

The market for digital cameras has grown dramatically for the past couple years. Like all new technologies, they are improving each month. The amount of picture storage is increasing. Finer resolution and larger image sizes are producing better and better quality pictures. One of the biggest advantages over the standard film camera is that digital cameras can perform digital signal processing, DSP. This processing can perform several functions such as decreasing the amount of space needed to store the images, correct for movement and focusing, as well as functions normally limited to software such as negative and grayscale.

This project will deal with the DSP function that corrects for colored ambient lights. This processing is called color balancing. Digital cameras observe the amount of light received regardless of the ambient light. The human eye, however, cancels out the ambient light. Our brains don't observe colors but color differences. A digital camera user would want the picture to ?look? the same as it ?looked? when the picture was taken. The problem for the digital camera designer is that the picture can be observed in a different ambient light then it was taken.

As a result, this ambient light correction needs to be performed by the digital camera so that the picture recorded will ?look? like the scene did to the camera user. This project will attempt do determing what type of processing does the Olympus C-2500L perform to accomplish this. This will also be extended to describe and implement standard digital camera color balancing algorithms. These algorithms will be analyzed to determine which algorithm the Olympus C-2500L performs.

As we mentioned, color balancing is an important feature of a camera and it directly effects the acceptance of output pictures of a camera. Therefore many color balancing algorithms have been developed for the digital camera to cancel out ambient lights. Based on their design assumptions, each of these algorithms is good at some light environments, but rather bad at the others. Thus choosing the right color balancing algorithm becomes one of the important design decisions of the digital camera. In the past few years, many projects and experiments have been done to simulate the color balancing algorithms and evaluate which algorithm is used for a certain digital camera. The algorithms have been evaluated in the past are Gray World, Perfect Reflector and Automatic color balancing algorithm. The cameras that have been evaluated include Sony, Olympus and HP Photo Smart. There are also some efforts have been done to evaluate camera color balancing output in XYZ space.

The problems in evaluating color balancing effects for cameras and algorithms include getting data of raw images, monitor gamma, camera gamma, resolution inconsistent and etc. The first problem can be solved by constructing the raw input images and taking pictures for the image shown on the monitor. The monitor gamma and camera gamma problem can be solved by evaluating monitor and camera gamma, then do the proper gamma or inverse gamma correction for input and output images. The resolution problem can be solved by picking the simple picture to make it easier to separate different color area, e.g. a color patch image. Based on these problems, the most often used image in the past work is Macbeth Colors Checker.

When evaluating color balancing algorithm, most of the past work only tested on a few images and drew the conclusions just on the results of a few cases. No one has evaluated color balancing algorithms over the whole spectrum and it is one of the major goals of this project.

trek@alumni.stanford.edu

lihui@leland.stanford.edu