EO Imaging Lab 2.1: Distortion

Hi, I am Greg Hollows and welcome to the Imaging Lab. This is Distortion. Distortion is one of those things that can actually be very difficult to deal with in a system, as you can see by the Fresnel lens I am holding in front of my face. We get some different adverse effects as I move the lens in and out. There are a variety of reasons for that from the image quality side of things. But it actually comes into play, what we are trying to look at here, is a distorted image that was created when we were moving that lens closer and further away from you. You can see that my face was blended out further and thus adding distortion into the system. Let's look at three common distortion types. In the slide that is shown, you are going to see a couple of different types of distortion that we are going to talk about today. There is barrel, pincushion and keystone specifically that we are going to discuss. Pincushion distortion is positive distortion. It is actually a change in positive magnification across the field of view from the center of the image. This essentially means that things are moving in a higher magnification range as you go out and creating what looks like a pincushion effect. Barrel is the opposite, that's a negative level of distortion, seen in many, many lenses, and it creates this kind of fish-eye, or curved effect, in the system. Much more information is being drawn in. The magnifications are lower than what is seen in the center and that's creating a distortion effect. We will get back to keystoning in a few minutes. When we look at these barrel or pincushion distortions, that's actually a change in magnification, as I mentioned, over the field. Now, there are some things to keep in mind with that. The first thing is that distortion is actually what we call a Geometric Aberration. Why is that important? The important part of that is that information actually isn't lost in the Geometric Aberration. It is just simply misplaced. Unlike some other types of aberrations that can happen in the system, having to do with color, it’s very hard to put information back together easily in that system. With Geometric Aberrations, since it's misplaced, if all the information is there, we know how much it's been misplaced by, we can simply remap it or measure it out of the system. Now, some things to understand there, is that distortion is not necessarily a linear property. It has some quadratic formulas to it and in some cases, even bends back on itself. If you look at the actual graphs that we have up now, you'll see two different types of distortion there. One is monotonic, and one is not. The one that is monotonic is continuing all in one direction, the other one is actually curving back on itself, to get a low distortion system across the field. As we look at the graphs, the vertical axis is actually the center of the field out to the corner of the field. The y-direction is actually the amount of distortion positive or negative that is actually in the system. Now, if we were to look at the graph where it's actually bending back on itself, if we were to take a distortion measurement somewhere in the middle of that curve, before it starts bending back, and extrapolated that all the way out, we would actually add more distortion to the edge of the system than is actually there because it is bending back to zero. It's important to know where these curves go when you're trying to factor it out of the system. There are a couple of ways to do that. First off is you can use the design data that the optical manufacturer might be able to give you to map that out. The second thing is to use some sort of testing target. Those test targets as can be seen in the image here will be made up of usually dots or squares of some sort, as we see in the image. We have black dots representing the real world or the actual image position through the lens. That's what it is actually seeing, and the red circles represent the real world position of those dots. We are able to do a calculation off of this across that field to remap the image, or just to calculate out that distortion that's in the system and get our measurements more accurate. This is for barrel or pincushion distortion. Generally, as a rule of thumb, you want to use a calculation method as opposed to a remapping of the image because it does take up an awful lot of process or intensity. Now, let's look at keystoning for a second, another type of distortion. Anybody who has ever worked with a projection system and has tried to do a PowerPoint presentation out on the field, has usually run into this. You get a table with a different height, a screen of a different distance, you put your projector down, and you get this keystoning effect of the image. And there is a nice little button on there that allows you to actually adjust for this and you can see your image shifting around. What that comes into play with that sort of distortion is, generally a tip or a tilt in either the image plane or the object plane of the system. We are getting different magnification levels in the system, with a standard lens at least. And as we go across the field, we are going to get narrow at one side, where the tip is in one direction and it's going to get wider in the other. This can really run into problems when you're trying to stitch images together. If you have any sort of keystoning effect, and there can be some amount in almost any imaging system because it is nearly impossible to get everything perfectly perpendicular between the sensor and the optical axis and the object itself, if you try and stitch these images together without taking keystoning into consideration, you can get a kind of warped image if you do each one of the stitching. It will also lead to measurement inaccuracy at the tops and bottom of the image potentially because you have different fields of view and your calibrations could be off. So, you need to take this into account and again, use a target of some sort to back this information out of the system. Some other quick things about distortion to keep in mind. Distortion is something we see every day. We see two to three percent distortion in our eyes and our brains are wonderful image processing devices and they have actually adjusted for this. If you are looking at a system though that is on a screen or something else and try to determine if there is distortion in there, you probably won't see it if it is at a fairly low level. Again, it is important to use software and some sort of measurement device or target to able to back this out and verify how much distortion you have in your system. Another thing that can come into play. Distortion can actually change and in many lenses, it does change as you change the working distance in the system. So, if you are going to go and use different working distances with different lenses, you can't use the same calibration all the time. As that lens changes its working distance, and you reset it for the system, the distortion will actually shift a little bit, so you should reconsider in re-calibrating and making sure that the system still functions to the level of reliability that you would like. That's Distortion. The next portion of this module is about Telecentricity. You can also click on any of the links shown on the screen to take you to another topic of interest.