Public Lab Research note


KAPvolution

by cfastie | February 25, 2016 00:43 25 Feb 00:43 | #12752 | #12752

Above: The ascent of KAP cameras.

The first camera most Public Lab mappers put into the air is a Canon PowerShot. These are cheap, lightweight, and easy to trigger remotely. The typical small PowerShot has a sensor that is 6 mm wide -- you could fit two of them on my pinky fingernail. There can be 16 million pixels on the sensor, but the small size limits image clarity and light sensitivity.

sensors.JPG
Above: The relative size of the sensors on the four cameras compared here. The actual size of the large one on the right is less than an inch across. Data from http://cameraimagesensor.com/size/#70,55,210,31,a.

I have been using a PowerShot S100 which has a sensor slightly larger than the typical small PowerShot, and I have been pleased with the results. This camera has lots of features that make it an excellent choice for aerial photography (see list here). So the S100 (also the S95 or S110) is an evolutionary step away from other PowerShots toward improved image quality.

After a couple of years of stasis, I have been pondering how to punctuate this equilibrium with the next step in KAP cameras. So I just won an ebay auction for a Canon EOS M. This "mirrorless" camera is not a DSLR (digital single lens reflex) because it lacks the mirror which provides through the lens viewing. It differs from point and shoot cameras like PowerShots because it has a huge sensor and interchangeable lenses. Mine has a "pancake" lens which is not a zoom lens -- it is an f/2 22 mm prime lens (an equivalent focal length of 35.2 mm). The combination of a sensor as big as in a DSLR and a prime lens has the potential to provide image quality substantially better than any camera I have flown with a kite.

4charta.jpg
Above: Specifications of the four cameras compared here. Data from http://cameraimagesensor.com/size/#70,55,210,31,a.

So I had to see if the image quality was actually better. I took the same photo with the four cameras in the figures above. This includes a typical workhorse PowerShot (A1400) and the fancier PowerShot S100. It includes the EOS M and also my Nikon D3100 DSLR with an ED 18-55 mm kit lens (f/3.5-5.6). Except for the EOS M, the cameras have zoom lenses which are wider than the prime lens on the EOS M. So I made two comparisons. I zoomed the zoom lenses so they matched the field of view of the EOS M and took the same photo with each. But aerial photos are typically taken with the lens zoomed out to its widest, so I also did that and took photos so that the same birch trees were at the right side of each photo.

FOV2k.jpg
Above: Two comparisons were made. The zoom lenses were zoomed to match the 35 mm (in 35 mm equivalent) prime lens of the EOS M (green). Photos were also taken with each zoom lens at its widest setting (blue). Focal lengths are all in their 35 mm equivalents. S100=24mm, D3100=27mm, A1400=28mm.

When aerial photos are stitched together into maps or panoramas they are typically displayed in viewers which can be zoomed like Google Earth, so users will often confront the limits of image resolution. Zoom lenses sometimes exhibit soft focus near the image edges when zoomed to their widest focal lengths. Zooming in just a bit can improve the edge to edge image quality, especially on inexpensive lenses. Edge quality is important when stitching maps or panoramas because the edges must stitch with the centers of adjacent photos and edges often become part of the final composite image, covering central parts of other photos.

EOS2marqee.jpg
Above: The two marqueed areas (blue) near the right side of the photos are compared below. This is the photo taken by the EOS M.

To better simulate aerial photography, I used very fast shutter speeds so the aperture was large (small f-stop). With wide open apertures, lenses produce poorer quality images, so this made it a tougher test. The larger crop of each photo reveals subtle differences in color and contrast among the cameras, but the clarity is not dramatically different unless you look very closely.

Far35cr4pan.jpg
Above: The larger crop of each photo taken with the lenses zoomed in to match the 35 mm (equiv) focal length of the EOS M prime lens. If you look closely you can detect some differences in image clarity. But if you don't need to look closely, there might be no reason to use one of the better cameras.

Examination of the closer crops of the photos (below) reveals more conspicuous differences in clarity. There is a cloudiness in the A1400 crop suggesting that an upgrade to the S100 can provide substantially better image quality. The EOS M photo is slightly crisper than either the S100 or the D3100 DSLR photo. This is probably due to the advantages of a prime lens over a zoom lens, and in the case of the S100, also to the bigger sensor in the EOS M.

Far35cu4pan.jpg
Above: The closer crop of each photo taken with the lenses zoomed in to match the 35 mm (equiv) focal length of the EOS M prime lens. Improving image quality proceeds clockwise from top left.

Zooming the lenses in to match the EOS M prime lens should improve image quality for two reasons: the zoom lenses could perform better when they are not at their widest, and the scene is magnified by zooming in. The photos taken with each lens at its widest focal length allow a comparison of how the lenses might be used for aerial mapping. As expected, this comparison reveals slightly more degradation of image quality in the three cameras with zoom lenses.

farwidecu4pan.jpg
Above: Close crops of the photos taken with the lenses at their widest focal length. The EOS M crop is the same as the one in the image above, but the others are from different photos.

Image quality varies among these cameras pretty much as expected. It seems that image quality is a function primarily of lens quality. Sensor size also plays a role, but the EOS M and D3100 DSLR have essentially the same size sensor and have noticeable image quality differences. This is probably because the EOS M had a prime lens and the D3100 had a zoom lens. The number of megapixels captured by the cameras (see chart above) is probably irrelevant. These results would likely be the same if each camera had an 8 MP sensor.

The results support the idea that taking mapping photos with a PowerShot zoomed to its widest can unnecessarily sacrifice image quality. Especially with cameras that have the equivalent of a 24 or 28 mm lens, zooming in just a bit can allow crisper photos. Some PowerShots have the equivalent of about a 37 mm lens, and zooming those might reduce the field of view too severely.

I think these results confirm my notion that the EOS M will improve the image quality of my aerial photos. These tests were done in bright sunlight with the ISO of each camera set to its lowest (slowest) setting. When the ISO has to be set higher to compensate for dimmer lighting, the large sensor of the EOS M will shine even more compared to the PowerShots. In fact, I will probably use the EOS M with the ISO set at 200 to 400 to allow the aperture to be smaller. This could improve image clarity without adding noticeable sensor noise. Using PowerShots with the ISO set to 400 or higher can produce conspicuous graininess in the photos.

The EOS M combines a nice big sensor with an excellent prime lens so it takes very impressive photos. Otherwise, it is a terrible camera. I would never choose it for regular photography. It lacks a viewfinder, and composing and focusing with an LCD severely compromises certain types of photography I like to do. It is intolerably slow to process photos and be ready for the next shot. Autofocus is also slow and unreliable. It lacks a flash and most of the common external controls of a DSLR, so settings must be made with the sluggish touch screen interface. This is the 2012 model and there is now an improved EOS M2, EOS M3 and EOS M10. All of these shortcomings of the EOS M are wonderful news because they have no effect on kite photography and allowed me to buy a used EOS M (with lens and shipping) on ebay for $173.00 (original 2012 list price was $799.99). That is just about $30 more than you might pay for a used PowerShot S100 (original 2011 list price was $429.99). Well worth it to get the improvement in image quality.

The EOS M weighs almost twice as much as an S100 (365 g vs. 198 g), so that is another price to pay for a KAP camera. I am still working on the best way to trigger the shutter remotely (CHDK works only on PowerShots). I hope to have more news on that front soon.


1 Comments

Magic Lantern seems to run on the EOS M and has an intervalometer. I can't recommend Magic Lantern enough. I've got an EOS 600D/T3i that I much prefer for most purposes over my wife's 70D which Magic Lantern doesn't yet support.

Reply to this comment...


Login to comment.