Is an 80-megapixel Imager Too Good to Be True?

Posted:  
Mon, June 11, 2018
Author: 
Shaina Warner  | 

So, you’re looking for an imaging system and you’ve just been pitched an 80-megapixel camera with a small price tag… Here are a couple of things you should know before buying in:
 

  • Did you know?
    • When a camera captures an image, photons of light are mapped to its CCD pixels. Pixels on a CCD can be different sizes. A small pixel has a smaller storage capacity for photons (called its “well depth”), while a larger pixel has more storage capacity. Because they can store more photons, a larger pixel capture images with greater differentiation between light levels (contrast ratios that make up a camera’s “dynamic range”) to enable more precise image analysis. All cameras capture images with a certain amount of inherent electron noise, at several electrons per pixel. Larger CCD pixels that capture more photons increase the ratio of true input (that accurately captures the photons reflected to create the image), to false input (that is electron noise).
       



      An example of how a large CCD pixel captures more photons, and—once filled—will provide more good signal compared to the amount of electron noise.

               

      Image with low image noise (left) versus high image noise (right).

      We call this ratio the “signal to noise ratio” (SNR) – where the signal is the amount of true input, and the noise is the inevitable, undesired electron activity. A large CCD pixel capturing lots of true input (signal) as compared to the amount of false input (noise) has a high SNR, while a small CCD pixel capturing less true input (signal) to false input (noise) has a low SNR. High SNR provides an image with accurate light and contrast variations, which is critical when using imaging systems for measurement and analysis.





      An image with low SNR makes it harder to parse the signal from the noise (top). An image with high SNR captures predominantly signal, which is easily discernable from the noise (bottom).

      Find another great example of SNR from Cambridge in Colour: http://www.cambridgeincolour.com/tutorials/image-noise.htm

      Consider this example: Imagine you are speaking at standard volume in a room. You are producing a signal that is meant to be interpreted by a receiver (listener). In a noisy room, your signal is harder to hear. Like with an imaging system, the point is to establish a scenario where there is enough difference between signal and noise levels for the intended sound (your voice) to be distinguished from other ambient, unwanted input (noise in the room).
       
  • What’s the problem?
    • CCD sensors, and their pixels, occupy physical space. The more pixels you put on a CCD sensor, the smaller the pixels must be. Alternatively, if you want to increase the pixel size in an 80-megapixel CCD (to maintain a high SNR, for instance), the physical CCD sensor size must also increase.
       



      Given the same number of pixels per CCD, as pixel size increases, the size of the CCD must also increase.

      It follows that you’ll need a camera with larger components to accommodate a larger CCD, and a larger lens to capture enough light to map to every pixel in the large CCD. Nonstandard components like these are much more expensive than standard components. In the 1920s, Leica set the standard in lens size, making 35mm the most common and widely-available camera format on the market. A lens of this size captures enough light to map to a limited physical area. In other words, a CCD can only be so big before the standard 35mm lens is no longer able to supply enough light to make use of every CCD pixel.




      For a CCD to fit the imaging area captured by a standard 35mm, the CCD pixel size must also be limited. Increasing the size of the pixels, without reducing the number of pixels, increases the CCD size beyond the imaging area of a 35mm lens (right). This means that some of the CCD area will go unused, and images captured by the CCD will not be full resolution.
       

  • What’s the solution?
    • To keep the cost of the camera down, the best bet is to choose standard camera hardware and lens, and therefore limit the CCD size. This means that the physical space allotted to us for photon-receiving pixels is also limited. However, there is one remaining decision – do we choose a CCD with fewer, but larger pixels; or do we use many, smaller pixels to receive photons on our CCD?
       
  • What are the considerations?
    • Resolution is key. Resolution is the number of pixels within the physical area of a given CCD. A CCD can maintain the same physical dimensions while increasing resolution – for instance, an 8MP CCD can be the same size as an 80MP CCD. What’s the difference between these two CCDs, then? As we’ve noted, to increase the number of pixels on a CCD of a limited size, we need to make the pixels smaller. At face value, an 80MP CCD within a camera of a standard size with a standard lens seems to be the best value for the investment. You get higher resolution without paying for the cost of a larger camera—great! However, each of pixels on the CCD of this standard-sized, 80MP camera is incredibly small. As we discussed, using smaller pixels means smaller capacity for photons and thus poor SNR. Although “80MP” would suggest better quality images due to higher resolution, if the pixels are small it really means you have a greater number of inaccurate pixels.
       
  • What is the optimal imaging system?
    • The optimal imaging solution depends on your application. For cameras used in personal-use devices—like your smart phone—a bit of image noise may be acceptable. For a camera used in measurement applications where precision is more critical, the difference between a quality component and a failure could lie in the details—or contrast levels—that are measured in the image. Radiant cameras are meticulously engineered to strike a balance between the number of pixels on a CCD and the pixel size. Obviously, we want more pixels to capture an image in finer detail (resolution), but we don’t want all of those pixels to be so small that they are messy with noise. We want larger pixels to increase the chance of getting a good, clear signal (since we know that noise is a smaller drop in the bucket of a larger pixel’s larger well capacity). However, we don’t want our signal to be interpreted by fewer pixels either. The optimal solution is a compromise for the highest resolution (more pixels) with the lowest noise (larger pixels).
       


      An image example of a high-resolution/high-noise image (left) versus a low-resolution/low-noise image (right). Neither is ideal for measurement—we need to strike a good balance.

      Bonus: Take a look at this camera comparison chart from Andor to see how the pixel size of a CCD contributes to dynamic range.

       

  • Takeaway:
    • If a low-cost, 80MP camera seems too good to be true, it probably is. It’s important to consider all specifications of a camera to ensure the optimal solution. Resolution doesn’t stand alone—it is impacted by several factors, most notably the physical limitations of the CCD, which may contribute to or reduce image noise depending on pixel size.
       
  • The best blend of specs:
    • A 29MP, 5.5μm-pixel-size CCD imager with high dynamic range using a standard 35mm lens provides the biggest bang for the buck in the image-based measurement world today. Incorporating CCD cooling reduces thermal noise yielding higher-quality images in a cost-effective solution.
       



      Image from above example, captured by a 29MP Radiant ProMetric® measurement system.
       

Capturing and analyzing measurements in extremely fine detail is needed to ensure illuminated components, displays, light sources, and surfaces are accurate to spec and continue to reflect the quality of your brand. Radiant has built its legacy on our imaging expertise. We have spent over 25 years refining our capability, from selecting the highest-performing scientific-grade CCD sensors to incorporating thermoelectric cooling for low image noise in our systems. Our ProMetric® imaging systems offer resolutions up to 29 megapixels with true, proven 64.1 dB system dynamic range, 1x1 binning without image averaging. Our dynamic range is measured based on the full system performance—it’s not simply the best possible CCD specification quoted from the sensor vendor. Our quality offers more detail per image (resolution) as well as more grayscale values (dynamic range) than competing systems, enabling precision measurements based on the finest visual contrast differences. With CCD cooling technology, we optimize the amount of captured image signal as compared to noise to ensure our systems are measuring only meaningful data.

Is 80MP imaging technology available? Yes! (And even greater for that matter.) But we know that an 80MP camera with small pixels sacrifices a good image, while an 80MP camera with larger CCD comes with increased hardware expense. At Radiant, we have engineered our systems with the intention of balancing image quality and cost-effectiveness, thereby providing the greatest possible value to our customers.
 



Radiant ProMetric imaging systems come in resolutions from 2-29MP, offering proven 61.4 dB (1x1 binning) and 73.4 (2x2 binning) dynamic range, and several lens options from 24-200mm, as well as wide-field-of-view and conoscopic lenses.



Image captured by a ProMetric I29 Imaging Colorimeter. High SNR enables precise uniformity measurement for all points on this illuminated display simultaneously.
 

To learn more about the balance of CCD size, pixel size, resolution, and dynamic range for capturing and analyzing images, read this technical white paper from Radiant, “Resolution and Dynamic Range: How These Critical CCD Specifications Impact Imaging System Performance.”

 
 
 
Share