Demosaicing and false colors test

There are other and more important uses for the ISO 12233 test chart. Measuring resolving power with it, is not really relevant in the real world. Of course, we all are fascinated by how much a new and exciting camera can resolve. But it is not as important as the demosaicing and false color filters that ALL cameras use to produce high image quality at very high frequencies – high Nyquist frequencies.

Image artifacts such as false colors, moiré and maze affect are a result of the Bayer system which as you know (or should know), distributes the incoming color spectrum into four components, so that each pixel can absolve only a third part of the spectrum.

After the image has been captured, camera makers need to reconstruct those 4 components (GB, B, GR, R) into one big RGB image. They do so by using the demosaicing algorithm.

The biggest problem with this algorithm and the Bayer system is that any data behind the Nyquist frequency can not be reconstructed. Consequently, many image artifacts including false colors, moiré and maze result from that limitation.

To make things a bit easier, camera makers place a special filter in front of the sensor called an Anti-Aliasing or Low pass filter. This optic filter reduces the frequencies that reach the sensor into the frequency range the sensor can sample and the demosaicing algorithm can reconstruct. However, the stronger this optic filter the softer the image and so, camera makers try to find a balance between the filter and the image processing capabilities to produce a sharp artifact-clean image.

This is finally where our test kicks in. When the demosaicing algorithm gets confused and cannot reconstruct the image, artifacts like false colors and moiré appear. And they DO appear in every single Bayer based camera out there, except that you don’t see it that much. That’s because camera makers mask those artifacts with gray blur. Some do it quite well and some don’t. Some have a very good demosaicing algorithm and some don’t. This is exactly what we try to evaluate with our demosaicing and false color test. Good demosaicing and false colors reduction filters produce a high quality image free from moiré and false color artifacts in the real world.

An image after demosaicing with no False Colors reduction
An image with False Colors masked with gray by the algorithm