Stitching for resolution

[johncox]

<p>I've sorta been under a rock for a while and haven't done any real post processing for a long time. If I took 7 or 8 shots handheld and stitched them together would I see an increase in functional image size (along the lines of how Olympus does their 40 MP JPegs)?</p>

[john_wheeler6]

<p>While you can do several things to improve the image in your scenario, it would not be along the same lines as the Olympus 40M jpeg (or Raw). In the Olympus case they clearly state that this needs to be on a very stable tripod viewing a still image. Vibrations have to be low so the mirrorless body has an advantage that might be match if you locked up on a body with a mirror.<br>

<br />Then they take 8 shots each moved by 1/2 pixel. Though they did not give details, the software must do a deconvolution algorithm to deduce what the luminosity and color of an intermediate (non-exiting) pixel would need to be.</p>

<p>So to replicate Olympus, you would need to know how to shift where the camera point by 1/2 pixel accurately 8 times with less noise than 1/2 pixel (hard to do even on a tripod) and have the software to do the deconvolution (someone could come up with that part no doubt).</p>

<p>So very doubtful you could duplicate that capability in the Olympus and have it be fruitful. Just my opinion of course.</p>

[Hector Javkin]

<p>If you were to shoot (on a firm tripod) with a longer lens, and use more frames to cover the same field of view, you would get much higher resolution. For example if, instead of shooting a particular scene with a 50mm lens, you were to stitch together four frames shot with a 100mm, you could theoretically get four times as many pixels. In practice, some overlap between frames is necessary, so the benefit is smaller. As John pointed out, this is not the same method that Olympus uses, but you will have an increase in resolution. Also, the overlap may have to be large, because lenses are sharper in the center of the frame than they are in the corners. You don't want too much loss of lens performance inside the frame. A variant of increasing the resolution is, instead of using longer lenses, to shoot with smaller format cameras.</p>

<p>For an easy way to look a the field of view of different lenses and sensor formats, you can go <a href="http://fcalc.net/online/"><strong>here, to f/Calc</strong></a>.</p>

<p>You might want to consider whether the effort is worthwhile. How large and detailed do you need your photos to be? With some of today's cameras 36x24mm sensor cameras providing 24, 36 and 50 MP, you can produce very high resolution photos. Of course, even more, at much higher cost, is available in medium format cameras.</p>

[Ed_Ingold]

<p>Some medium format cameras use a similar technique, where the sensor (or Bayer filter) is moved by one pixel rather than combining adjacent pixels (RBGG) and interpolating colors.</p>

<p>From a practical point of view, it is probably more effective to use a longer lens with the camera in "portrait" orientation, and stitch frames together using a panorama program like PTGui. Typically, you want a 25% overlap between frames, so it would take 3 vertical frames to constitute a 3:2 aspect ration in the results.</p>

<p>The downside is that with an high resolution camera (24+ MP), file sizes become very large, and may tax your computer or even exceed limits of the software. At one time, using PTGui, I had trouble stitching 16 MP (MF digital) frames together. Since then the processing capacity of the software has been increased. I can stitch 5 or 6 vertical frames from my 42 MP Sony. With a frame size of 7900x5300 pixels, I'm not inclined to do that very often. The plus side using the A7Rii with IBIS, I don't absolutely need to use a tripod.</p>

[MathewDH]

<p>The digital images already have a lot of detail if they were taken correctly in RAW. I prefer to not do the post-processing until the images are stitched together. So do the minimal RAW processing and output as uncompressed 16 bit TIFFs.</p>

<p>When stitching the images, as in PTgui, you can specify the final pixel size of the stitched 300 DPI image. Which allows you to make the image size large enough to allow cropping down to the final print size. The advantage is you get all the detail available for larger print sizes. If you stitch to a smaller size and then resize to a larger image, the quality of the larger imager will be of a lesser quality. You are better off stitching images for prints of a larger size, then resize down if you want smaller prints.</p>

<p>Then do the final processing to bring out the best qualities of the final stitched image.</p>