Conclusions

• Gray world color balancing does a good job on images with a wide range of colors (which is true for many natural images), but will create false colors when used on an image that is predominantly one color.

  • Additive gray world works better than multiplicative, which will create false hues of deficient colors

• Illuminant-based color balancing has similar results to gray world on images with a wide color range, but works much better on color-deficient images. However, it can only be used if the illuminant is known.

  • The multiplicative method tends to work better than additive, because the scaling factors are based of a white image and will not have to problem of over-scaling that gray world has.
  • Attempts to create an algorithm to guess the illuminant from the image failed because the color range in natural images is much greater than the color range imparted by different illuminants

• For brightening, a method that brightens dark pixels more than bright ones tends to work because it doesn’t push the image into saturation

  • Gaussian brightening does a decent job of bringing the subject out of a blown shot, but a significant amount of tweaking is necessary to brighten the right range of digital values an appropriate amount