DE razor geometry. A system for measuring aggressive razors.
- Thread starter Slock
- Start date
I don't mean to toot my own horn, but I was the one who pointed out a photo with much less skew than this that prompted @Rosseforp to re take his pictures. I said I was sorry for being picky and he was happy somebody caught it. My only point is that I don't see skew with the images in question. That leaves distortion or manufacturing. Your post helps show the concept of skew throwing off measurement, so don't take this wrong please. We are trying to put together a method that everybody can replicate at home so we can compare images. The photos from @JPO show that phone cameras may not be ideal considering that the macro lens cost the same as your whole microscope.
Then it makes more sense
So maybe my karve razor is just not machined with the same accuracy as the Tatara razors.
I has taken me a while, but I think I have an explanation how there could be so much distortion and how that could affect the image. I am just not used to seeing this in macro photography.
Phone cameras have wide angle lenses mostly, even if they say telephoto. I don't want to explain optical formulas, that will get us sidetracked. Wide angle lenses produce barrel distortion, some more than others. Without knowing enough about the macro lens, I am going to assume it is Gaussian. To explain the distortion, you would need to visualize the image wrapped around the top of a globe. But the optics are stretching the image before it is produced. The optical characteristics of a lens are complicated and not completely spherical or they would be a fish eye lens.
It is possible that the center of the image, the center of the razor, is relatively flat. But, the safety bar and blade edge are bent by the optics away from the camera. This would produce the same phenomenon as having the razor skewed. The blade is less bent by the optics, being closer to the center, but still some. The safety bar is bent more, being further out. This might be affected by where the camera is focused as well.
The photo of the Tatara looks different and the measurements are more closely aligned with the manufacturer. Whether this is how it's photographed or better manufacturing, I don't know. As I said above, I partially retract my anger at Karve's misleading specs. They are clearly off, but how much remains undetermined for me.
Phone cameras have wide angle lenses mostly, even if they say telephoto. I don't want to explain optical formulas, that will get us sidetracked. Wide angle lenses produce barrel distortion, some more than others. Without knowing enough about the macro lens, I am going to assume it is Gaussian. To explain the distortion, you would need to visualize the image wrapped around the top of a globe. But the optics are stretching the image before it is produced. The optical characteristics of a lens are complicated and not completely spherical or they would be a fish eye lens.
It is possible that the center of the image, the center of the razor, is relatively flat. But, the safety bar and blade edge are bent by the optics away from the camera. This would produce the same phenomenon as having the razor skewed. The blade is less bent by the optics, being closer to the center, but still some. The safety bar is bent more, being further out. This might be affected by where the camera is focused as well.
The photo of the Tatara looks different and the measurements are more closely aligned with the manufacturer. Whether this is how it's photographed or better manufacturing, I don't know. As I said above, I partially retract my anger at Karve's misleading specs. They are clearly off, but how much remains undetermined for me.
I added this segment and got 0.96
Would taking a picture of a grid like graph paper show the type of distortion and allow us to measure the deviation?
Last edited:
You caught on to where I was going. The short answer is maybe not! I spent some considerable time thinking about those Karve images, they bothered me so much. This is really getting at why we need RAW images. But as I said, I concede that the microscope delivers the goods. I can't really go into the whole optics, I am literally giving myself a headache thinking about this. From a photography point of view, not having control of every single variable and having to guess what the camera is doing under the hood is not a good place to be.
I will give some insight into my madness. The reason we may not see distortion with a grid is because of the way the image processor assembles the image from the sensor. We have no idea what sort of correction it is applying nor can we reverse it. The traditional way photographers test their lenses is with a brick wall, but I don't think that is helpful at close distances. There are optical tools that create a profile of the lenses distortion, but those are way beyond our purpose. I cannot remember if closer focusing reduces or increases lens distortions and my books are in storage. A lot of this is not intuitive.
I think we can all see that this image has the razor tilted back slightly, exposing the underside of the top cap. I still don't believe that direction is causing the problem. However, looking closely at the underside of the top cap gives me concern. There appears to the slightest curve to the edge, but it is near the center of the image. What is more concerning is whether the safety bar at the edge of the image is being stretched to the left and the top cap pulled up. This is what I mean by staring at it trying to figure out what the distortion originally caused and the image processor interpreted and rendered. It may be useless trying to get a measurement from this particular one, but I can't help trying to figure out what is going on.
Probably the easiest way, with enough resolution, to reduce this is to photograph further out so the target area is no more than the middle third of the image. Then digitally zoom in and see if you have enough resolution. I discovered this last week working with my mobile phone, but I was stymied by lighting problems.
A visual check with a steel ruler next to my tatara indicate to me that the cap is closer to what the photos show then the specs from Tatara. I have emailed them.
This is from their web site: My eyes tell me this is not correct. It is much closer to 1 mm.
Original Mamba .53/Kai (3). Appropriate angle, Bob's your uncle. Glad I kept it.
Sent from my Pixel 4a using Tapatalk
Sent from my Pixel 4a using Tapatalk
Lens distortion, and the post-processing that corrects it is fascinating stuff. I've been reading a bunch on the Imatest website. They make the software that evaluates and measures lens distortion for lens manufacturers and camera manufacturers. Here is a link that has a nice video overview of lens distortion.
Distortion: methods and modules | Imatest
www.imatest.com
The software license is $12k, which is over my budget, but the methodology for diagnosing distortion with the $12k software is of definitely of interest, and it's free. Take a picture of a grid. If lines aren't straight, there is distortion. It doesn't matter if there is post processing, or algorithms, or digital magic. If a picture of a grid of straight, parallel lines becomes curved, or cease to be parallel, there is distortion.
It probably makes sense to see how bad the distortion is before using a lens to take measurements. If it's as bad as the pic below, which it may or may not be, it doesn't make a lot of sense to invest time taking measurements with it. All macro lenses cause pincushion distortion, the question is whether this particular lens has distortion large enough to affect measurements, and over what distance. A picture of a grid or checkerboard pattern with the macro attached would be ideal, but a picture of a ruler would be informative.
These photos come from reviews of the Apexel Professional HD 12X/24X on Amazon.
Last edited:
You can't blame me for pushing you down that rabbit hole! I deliberately didn't link any websites like that, but there are more. That last image really shows my "image wrapped around a globe" analogy nicely!
I know @Rosseforp is busy with his move, but I get a feeling eventually he is going to come in here with brand new Comparator images and drop the mic on all our feeble efforts! 
Keep in mind also, that the distortion is proportional on the top and bottom because the lens is a cross section of a spheroid. So let's pretend this is a picture of a razor. I just cut the ruler pic above in half flipped it over. We have a razor cap on top, and razor baseplate on the bottom. Let's measure the gap. My fancy ImageJ software let's me measure with precision out to 1 one thousandth of an inch.
I measure exactly 2.133mm. The actual gap is 4mm. This is what I meant when I said in an earlier post that ImageJ "allows me to take inaccurate measurements with incredible precision".
I measure exactly 2.133mm. The actual gap is 4mm. This is what I meant when I said in an earlier post that ImageJ "allows me to take inaccurate measurements with incredible precision".
Last edited:
This isn't correct and I didn't want to have to get into it. The image a lens projects is larger than the sensor. We have zero guarantee where manufacturing tolerances put the projected image. It could be symmetrical diagonally! It is possible for the sensor, by design, to only occupy the center third of the projection because the aberrations in the lens are too extreme at the corners. Then you have manufacturing tolerances which may randomly put 10 different samples at 10 different places on that projection. Because we are talking about cell phone cameras, I believe the whole sensor is 3 or 4mm! It is very much the same affect as your centering device if you are just slightly off. In fact one way to visualize this is to get a set of "Art" lenses for a home made pin hole camera and see, crack the lid slightly, where they project an image on the back. If you wiggle the lens slightly and the projection moves all the way to the corner, don't be surprised! You are also going through a clipped on lens which is not exactly a precise alignment. And it has it's own optics with their tolerances. I am sorry for the bad news, this is why I conceded that the microscope is the current best solution. I don't like unknown unknowns.
Last edited:
Interesting, and potentially problematic for sure. Is this not something we can test empirically however? If we take a photograph of 600 squares that are each precisely 1mm and take a photograph of it, and the photograph shows every square at all points in the image to be exactly 1mm, I think that demonstrates the absence of distortion. If we are able to do this 10 times in a row, have we not demonstrated repeatability? This stuff goes way over my head and I don't properly understand it.
This is the test I did earlier in the week with my microscope just to satisfy my own curiosity, not knowing anything about the subject. It just seemed like a logical test. Is this incorrect?
That doesn't look too bad at all! I'm measuring about 0.07mm deviation at the centerline of the 10mm span.
While this is very encouraging, a single line doesn't necessarily rule out distortion. Even in the totally messed up distorted picture with the Apexel 12x/24x, there are straight lines in the image. It gets weirder the further from the center you get. In some places there is barrel distortion, in some places there is pincushion distortion, and in some places there is mustache distortion, and in some places there is almost no distortion. That's why a picture of a grid is so helpful. Take a look at the picture below. Look at the red lines. Crazy.
Fortunately, a 1 mm grid is very easy to produce. I just made mine in microsoft word and printed it. It takes 60 seconds. Here is a vid on how to automatically make a 1mm grid pretty much instantly.
https://www.makeuseof.com/tag/make-graph-paper-microsoft-word/
I think with your microscope, it is correct. I remain uncertain about phone cameras.
I can adjust the distortion in affinity photo. There is suppose to be 2% with my lens. However i am not sure what happens when it is combined with the phone lens.Interesting, and potentially problematic for sure. Is this not something we can test empirically however? If we take a photograph of 600 squares that are each precisely 1mm and take a photograph of it, and the photograph shows every square at all points in the image to be exactly 1mm, I think that demonstrates the absence of distortion. If we are able to do this 10 times in a row, have we not demonstrated repeatability? This stuff goes way over my head and I don't properly understand it.
This is the test I did earlier in the week with my microscope just to satisfy my own curiosity, not knowing anything about the subject. It just seemed like a logical test. Is this incorrect?
Can i maybe use this type of mm grid at the same distance as the top of my razor and calibrate my distortion filter? Just an idea.
