Wednesday, January 18, 2017

Comparison of inexpensive digital microscopes, part 3

I promised that my review of inexpensive digital microscopes would be a trilogy, and by golly, that's what it's gonna be. In part 1, I looked at the Celestron USB microscope. In part 2, I had a quick peek at a lens that turns your smart phone into a microscope. In this blog post, the third part of the series, I look at a microscope that I just purchased from Opti-TekScope. It sells for $89.

The Opti-Tekscope in the stand (left), and frees ranging (right)

I think this microscope is the greatest thing since the invention of the microwave DVD player. (I watched all 53 episodes of Downton Abbey in 37 minutes.) The microscope gives me very nice images at a resolution smaller than my two previous microscopes, and the stand is solid. 

If I can interject a techno-linguistic question for the moment, I have wondered about the proper terminology when you compare the resolution of two microscopes or images. I think it is common to say "higher resolution" when you refer to an image that has a higher magnification. But technically, resolution refers to the tiniest discernible feature. I could speak of being able to make out a dot that's 10 microns in one image and 100 microns in a second image. In common terminology, we would refer to the first image as having higher resolution, but that runs counter to the size of the numbers.

This is the kinda stuff that keeps me up at night.

Selfies

You get a new camera, what's the first thing you do with it? Here is a series of selfies that I took with this microscope. I think this is pretty impressive. I mean the images. Of course my beard is pretty impressive. I can almost use this as a webcam, and alternately, I can also tell whether my barber did a decent job trimming an individual hair of my beard.

Proof that I still have brown hairs in my beard
   
The eyes have it

Below I show a comparison of an image from the three microscopes I have looked at so far. How to characterize them???


Subjectively, I would say that the three images have similar resolution. The Opti-TekScope image on the far right has a slightly reddish cast - the magenta dots seem more predominant. The Celestron image on the far left looks a bit washed out. But overall? I wouldn't complain a whole lot about any of them.

Except if I started getting picky. As I said in my last blog, the image at the far left is not just washed out, but also a bit blurrier than the one in the center. But take a real close look at the image from the cell phone. Look on the left edge of the eyebrow. (See the circled area in the cropped image below.) There are some white splotches in there that look very suspicious. You don't see those white splotches in either of the other images. For the time being, I am going to explain these as artifacts in my cell phone camera. Yet another thing for me to investigate. That's gonna keep me up at night.

Are these white splotches spies on a covert mission?

But there's more... The aforementioned comparative study of the model's conjunctiva did not take full advantage of the Opti-TekScope. It's barely warmed up! How about this image comparison?

Now we got a gosh darn good image of the nevus!

It is clear from this set of images that the resolution of the Opti-TekScope is way more better than the other two. It would be ok to use the other two images to measure the size of the halftone dots, but you can truly see the details inside of the halftone dots with the OptiTeckScope.

Pixels on my KindleFire

In the previous two reviews, I had a look at Wikipedia. Not the website; the icon. I looked at the Wikipedia icon on my KindleFire. As you may recall, the Celestron gave me a usable image of the pixels in the display, but the cell phone microscope was unable to focus properly. How about a side-by-side image of the Celestron and Opti-TekScope? I think it's clear that the Celestron image is blurrier.

Which photomicrograph of a Wikipedia logo would you invite to a Starbucks date?

You may be wondering... Why do I point a microscope at a display? Teaching. I want to teach the idea that practically all the displays we look at (TVs, computer monitors, cell phones, laptops, and tablets) are made up of three colors of dots. All the colors that we see when watching a microwave DVD version of Downton Abbey are created by mixing the light from red, green, and blue pixels. I have given the "let's look at the pixels in a computer screen" demo in color classes about eleventy-seven times. (Sounds like something good for a blog post or two.) I think I finally have a good way to demonstrate the effect.

I think this clearly demonstrates that the resolution of the Opti-TekScope is more better than the Celestron.

The last stand

Finally we come down to what holds the whole thing up. If you recall the Celestron microscope, it has a stand that is adjusted via the articulation of three ball-and-socket type joints. I was not happy with it. In fact, at 3:00 AM this morning, when I finished obsessing over the correct meaning of the adjective higher in phrase higher resolution, I decided to start fuming about how hard it was to position the Celestron. Whatever. It was a good reason to get up an raid the fridge.


But the stand that comes with the Opti-TekScope works just like I would hope a simple microscope stand would work. It moves up and down, without messing up the direction the scope is pointing. By much. Focus is adjusted with a knob on top. The knob is gentle enough to not spoil the pointing of the scope. By much. 

I'm not going to claim that the stand is a work of art, or that it looks like it would withstand a nuclear detonation at 50 feet. That part I didn't test. But the stand is quite functional.


I have looked at three microscopes so far. Of the three I have played with, this one, the Opti-TekScope is by far my favorite. Come to think of it, it is also the most expensive. But for $90... I think it's a good deal. You should buy one for all your grandkids.

4 comments:

  1. Great, John. Thanks. I've had a cheaper knockoff for quite a while.
    I hope people realize this is not a purely optical scope, but rather lenses and CCD camera.

    The Opti-Tekscope does have better resolution color. I am curious, would you use a term better "color resolution" and how would you define that? Is it meaningful and would it be defined like DeltaE in perceptual space?

    ReplyDelete
  2. I can confirm that the optical part of the Opti-TekScope is a single, multi-element lens and a CMOS detector, with a screw mechanism that adjusts the position of the detector relative to the fixed lens. I would contend that, with a computer and software, this is functionally equivalent to a purely optical microscope, so I am not sure what your distinction is... Other than maybe "Warning - This product requires a computer."

    "Color resolution"? Hmmm... I think the big issue here is that, to the best of my knowledge, there are no RGB cameras that meet the Luther-Ives condition, which is to say, they have a spectral response that can be directly translated into the spectral response of the human eye. As a result, there are two instrumental metamerism issues: There are spectra that the eye would see as exactly the same, but a camera would give different RGB values to, and conversely, there are spectra that the camera would see as exaclty the same, but the human eye would see differently.
    http://johnthemathguy.blogspot.com/2014/07/rgb-into-lab.html

    A major printer I know of uses high end cameras in their photo studios, but spends $8 per image for someone in India to do color correction for every images that it prints.

    Another complication is that there are layers of software and hardware in between the native RGB values seen by the detector in the camera and the final display of the image on your monitor.

    I don't know if anyone has developed a standard metric for measuring this, but it would have to include the RGB values coming out of the camera, and whatever else that is done to convert this to CIELAB. Then it might make sense to apply something similar to the metameric index, which all comes down to deltaE as you suggest.

    ReplyDelete
    Replies
    1. Thanks. I looked up Luther-Ives. It makes sense and is as you said. I was looking for color precision, more than accuracy. Thus if the camera's channels each have 12 bits rather than 8 or fewer bits, its color resolution, or what I mean - its ability to distinguish two close colors - would be better. Now suppose the spectral responses of the three channels overlap on the 12-bit one, whereas they are distinct on the 8-bit one. Now which one has better "color resolution" as I defined it? How would one quantify this or would that be the metamerism index you mentioned.

      Delete
  3. Thanks for your continued interest, Mitch.

    Precision (defined in this case as bit depth) is only part of the issue. My experience with digital cameras from a while back is that repeatability is the bigger issue. Anything beyond 9 or 10 bit digitization was pointless. You just get more bits of noise. This work was done with a detector with 10 micron pixel size, which is large compared to cell phone cameras. Smaller physical pixels on the detector --> smaller full well capacity --> lower SNR --> less repeatability.

    And who uses a single pixel to measure color, anyway? That makes it all the more confusing. How many pixel do you average?

    Your second point about overlap makes it even more confusing. Consider taking your thought experiment to an extreme. Say we build a camera with a UV channel, a green channel, and an IR channel. Excellent separation, no overlap, and really high entropy (small correlation between the channels). Does this hypothetical device give us fabulous color resolution?

    It gives us a wonderful ability to discern one object or light source from another, but it also has a "wonderful" ability to find a big distinction between two colors that look exactly the same to the eye, but differ wildly in regions of the spectrum that we can't see. And it will miss any interesting stuff going on in the red and blue regions of the spectrum - that we can see. I'm gonna say that this camera has lousy color resolution.

    We make the hypothetical camera less extreme -- one might think that a sensor with something like Status T density response (which has a red, a green, and a blue channel with very little overlap) would have very good color resolution.

    But...

    The L and M cones (which we loosely call "red" and "green") have considerable overlap. This is the reason why we perceive a very large hue shift between 580 nm (greenish yellow) and 610 nm (reddish orange). If a device had high resolution but little overlap, it would incapable of discerning these hue shifts. And I would say that this would limit its color resolution significantly.

    ReplyDelete