In-camera Color Processing from a Mosaic Color Filter Array (e.g. Bayer)

Posted by Gretchen Alper on Thu, Jun 9, 2011

For certain imaging applications, a full color image is required for accuracy, particularly with outdoor applications, such as traffic systems (license plate recognition), surveillance, and situational awareness.  Adimec’s color cameras can include in-camera color processing which we developed ourselves to be optimized for automated adjustment to ever-changing outdoor conditions. 

Color processing is optimized to bring human color perception from a screen-image as close as possible to color perception when looking at the actual scene.  Adimec’s color cameras utilize single chip digital image sensors with a Bayer CFA for cost and response reasons. The filter pattern is 50% green, 25% red and 25% blue.  (This is also discussed in our previous blog post).  With Mosaic Color Filter Array, color processing involves 2 steps:  demosaicing and then color recalculation.

Demosaicing – filling the Bayer gaps

With a Bayer pattern sensor, each pixel collects data for one of the three Bayer primary colors (RB, GB, or BB). Alternatively put, the image consists of three subsampled color images (RB, GB, BB).  For each of the three, the gaps are filled by interpolation – this is often called demosaicing.  Inadequate demosaicing is one of the main causes of artifacts in color images, particularly near sharp edges.

Adimec created its own complex interpolation algorithms for successful demosaicing.  After the interpolation the data is tripled as there is now an R, G, and a B value for each pixel.

Color Recalculation

The demosaicing results in RB, GB and BB values for each pixel, but the resulting image may still not appear exactly as the human eye would view the object directly. 

Image sensors have a different spectral response (sensitivity per color) with respect to the human eye Figure 1.

Adimec color cameras spectral response 

Figure 1 Spectral response CCD (dotted) vs human eye (straight)

This means additional color recalculation is required to represent the color of the scene as it would be seen by the human eye. 

Adimec has created proprietary color recalculation algorithms which also include controls for saturation, protection from blur by interpolation, contrast enhancement and gamma.

The end result is a crisp image with accurate color reproduction of the actual scene (Figure 2 for comparison).  Automated in-camera functionality allows for consistent images regardless of ever-changing outdoor conditions.

Adimec color processing example

Figure 2.  Color image after just demosaicing (left) and final color image after applying color recalculation algorithm (right)

The in-camera color processing can also be disabled for the user to access the raw Bayer data if needed.

For more information on color processing from Adimec, please click here for our free technical brief.

Topics: Image Quality Improvements

Previous blog:

Next blog: