View Full Version : Problem with BY8 Debayering

February 1, 2011, 01:02:19

Firstly I am using a DBK 41BF02 With Imaging Control 3.1 and IC Capture 2.1 on Windows XP SP2.

When using BY8 I seem to be having a problem with bright red objects/colour. When the red channel saturates it introduces a pattern, which is only seen when using BY8, not when using UYVY, leading me to believe its something to do with the debayering on the PC.

The pattern doesn't change when the debayering algorithm is changed, although the extent of it can change a little (say on a curved object the point where the pattern starts can change).

The various colours in this test chart show that its only the red that seems to be a problem. With the bayer system green is special, but blue and red should be treated the same, so why do they perform differently here?



Sorry these are JPGs but the BMPs were nearly 6 meg each. These are minimally compressed so the pattern I mean is still easily visible.

This is a problem for my application as the objects we need to image are strongly backlit, leading to a pattern on their periphery.

Any thoughts?


Stefan Geissler
February 3, 2011, 15:26:37

First of all, I do not know, why there are differences. However, BY8 is calcualted in the computer while UYVY is running in the camera itself.

However, please try the latest driver from http://www.theimagingsource.com/support. Its version

February 3, 2011, 19:14:50
Hi Stefan,

Thanks for that, the problem is still there with that driver. Although I notice now that the pattern changes with the 3 different debayering algorithms. Lines with Nearest Colour, a grid of dots with bilinear and a maze-like pattern with edge-sensing.

Are you able to replicate this problem Stefan?

Stefan Geissler
February 4, 2011, 09:59:52

Lines with Nearest Colour, a grid of dots with bilinear and a maze-like pattern with edge-sensing.

This is normal and a results of the bayer alghorism. At least, the mazelike pattern.

According the other parts, I have no idea right now. I can only forward this to the programmers.

February 21, 2011, 22:49:32
Any advances on this? I recieved a DFK 51BG02.H today that displays exactly the same issue, and considering this camera only has BY8 available (no UYVY) it has become a problem!

Stefan Geissler
February 22, 2011, 10:15:34
First of all, you have set the bilinear alghorithm, the UYVY format uses edge sensing, thus the results are different.

The reason is, as far as we know, a result of the smear effect of CCD sensors. The reds are extremely bright, thus they smear in the greens in the next line. Therefore you get bright green columns which results in the pattern you see, when the sensor is over exposed.
The debayering itself seems to be correct. Therefore the effect is only visible in the reddish fields. However, if you had extremely blue light, then the effect is also visible on the blue fields.

Therefore, the problem is caused by the sensor itself and thus you unfortunately must live with this, except you do not overexpose.

February 22, 2011, 19:22:24
How come the artefact shows on the Edge Sensing algorithm when BY8 is selected but not when UYVY is selected? Does the processor on the camera perform some additional processing or filtering?

Stefan Geissler
February 23, 2011, 08:05:29

Please understand, that I do not publish internals of the firmware.

September 6, 2011, 02:10:38
I found another interesting behavior with this problem, it is only present when a particular camera is set to its highest frame rate. I have tried several different Gig-E cameras of different resolutions and they are all the same.

Could this be something to do with the pixel clock? Is the pixel clock changed with the frame rate or does it remain fixed to the setting that gives the maximum fps?

Stefan Geissler
September 6, 2011, 09:15:34
The maximum frame rate does kind of overclocking the sensor. This can result in some side effects.