[Home]

Camera's Use of Center in Color Balancing

Cameras are designed to minimize power usage and decrease delay. A technique which would help optimize these camera properties would be to decrease the amount of processing used during color balancing. This can be accomplished by only analyzing the center of the image to perform color balancing.

With Center Without Center

This can be tested by keeping the background constant and varying target colors outside the center. If the background color does not change when the targets change, the color balancing analysis is limited to the center. If the background color does change, the camera uses the target portion of the image in its analysis. This change, or lack there of, can be observed by ploting the difference as the target colors change. The background colors were obtained from the output images (after gamma correction) by averaging the color over the four corners (Shown here to the right).

Common sense tells us that the colors are going to change because of the inaccuracy in the measurement. So what level of change would be significant enough to conclude that the analysis is not strictly limited to the center? This can be determined by also observing the changes in the background color when the center target is changed. This will causes changes that are known to be a result of the center changes.

The following three plots show the changes in background color with and without a center for the three colors, R, G and B. These plots were generated using two scripts, CompareImagesCenterNo.m which compares the images without the center to the originals and CompareImagesCenterNo2.m which compares the images with the center. The data for the images without the center is stored in centerNoDifferences.mat. The data for the images with the center is stored in centerNoDifferences2.mat.

Red Color Test Green Color Test Blue Color Test

These images show that the camera's analysis for color balancing is not limited to just the center of the images, but is performed on the entire image.

trek@alumni.stanford.edu

lihui@leland.stanford.edu