When choosing a piece of hardware for our system we tend to take spares for any calculation tolerances we had in our technical budget or for later generations of that product and this basically in order to save the need to search and integrate different hardware afterwards. So, if we calculated we require 10GB disk space, we integrate 20GB disk; 20 I/O pins will be 30 I/O pins; 2 sockets of USB3.1 we order 4 sockets of USB3.2; all of our algorithms are in grey-scale mode and VIS monochrome camera was chosen and we integrate a color camera. Wait! What??? Color camera?
The purpose of this post is to demonstrate that unlike USB3.2 vs USB3.1 for example, a color camera is not an extension of a monochrome camera even if both camera sensors are based on the same sensor backbone. On the surface, it may seem like sensor XYZ-monochrome and sensor XYZ-color are the same however a color sensor has a pattern of color filters on top of the raw pixels which affects its optical performance.
A quick disclaimer and a spoiler: In this post I will clearly show that the monochrome camera has some advantages over the color camera. It is not my intention that from now on everyone will use a monochrome camera but only to make sure that when a color camera is chosen to be integrated into a system it will be chosen for the right reasons. Our choice for color camera will usually be when the final image in VIS eventually has to be displayed to a human: cameras, cell-phones and so on.
Head to Head Sensor Comparison – Color and Monochrome Cameras
To better show the differences between the two types we will take the following two cameras: IDS U3-3880LE-M and IDS U3-3884LE-C which are essentially the same camera (the latter is a non-packaged board level camera version) both with the Sony IMX178 with a pixel size of 2.4um, the former with the monochrome version of the sensor and the latter with the color version. The cameras were paired with a Computar V3522-MPZ 35mm lens. Both cameras and lens were provided by courtesy of OpteamX.
Sensor Sensitivity (Quantum Efficiency)
We will begin with a simple comparison of the quantum efficiency curves of both sensors, taken from Sony official datasheets:
It is easy to notice that in the VIS range the objective quantum efficiency per pixel is higher in the monochrome camera which, of course, makes sense– a non-filtered optical path will have more light throughput than a filtered optical path.
Another important note regarding the color sensor is the NIR range sensitivity: there isn’t any! If your application needs anything more than 400-700nm a color sensor may be problematic.
Image Resolution
The image resolution we experience is a function of the optics (lenses, mirrors etc.) and the sampling of the wavefront in the image plane which is a function of the image sensor of the camera. In a monochrome sensor all the pixels experience the optics and photons in the same way.
However, in a color camera the case is a bit more complicated. Let’s take a look at the filter patten below again taken from Sony website and also called Bayer filter. The filters are of red, green and blue colors (RGB) and we see that in this particular filter structure the red-filtered pixels are only 25% of the pixels, which in turn means that we sample only 25% of the image area in red. This means, in terms of sampling frequency, that for each color on its own we do not have an orderly sampling as opposed to the monochrome sensor and each color will have lower sampling resolution and hence lower sensor MTF than the monochrome sensor. Since the final image we see is a combination of the 3 colors, the immediate result will be lower MTF than the monochrome sensor. So, when the sensor sampling is the limiting factor, a color camera will experience lower overall resolution.
This explanation will hold for each pattern of filters and not necessarily specifically for just this one.
Experimental Data
To show what everything looks like in the real world, a USAF 1951 resolution target image was taken with both cameras in f#4.0. A white multi-LED illumination was set constant over all the images taken. See images below of the relevant resolution groups 4 and 5:
The images were taken with the same exposure time and clearly the brightness of the monochrome image is higher than that of the color image (an average of 193 grey level vs. an average 149 grey level). BTW, I assume I could have done a better job with the color camera’s white balance but that’s irrelevant to our point.
Furthermore, without going into too much image analysis, it is clear that we see more clearly the resolution targets in group 5 in the monochrome image. How much clearer? Around X3 in MTF contrast. See below the analysis of group 5-3. The first 2 graphs are line histograms of the vertical targets in group 5-3. The bar-chart shows the MTF (contrast) result of each color vs. the monochrome sensor:
A quick note: a common misconception when looking at the USAF resolution targets is that the color image is blurrier because it is susceptible to the lens chromatic aberrations. These chromatic aberrations are not eliminated in the monochrome camera though the monochrome camera is susceptible to the very same chromatic aberrations. When we take the exact same lens, the chromatic aberrations are the same in both cameras and so will be their effect on the resulting image resolution. The best way to see it will be taking a lens + camera combination where the camera sensor sampling is not an MTF limiting factor. In this case the manifestation of any color aberrations of the lens will impact the final MTF the same in both cameras.
To summarize, the answer is a definite no, the color camera is not an extension of the monochrome camera. When choosing a color camera for your application make sure to note the decreased sensitivity and optical resolution as well as the algorithmic overhead of turning a color image to a grey one and add them to your technical budget calculations. These differences are sometimes so profound to the point of systems that have monochrome cameras with filtering of the illumination which have better optical performance than a white light illumination and a color camera for the same setup for example – telescopes. Make sure to choose color camera only when you really need it.