0

The goal is to reduce noise.

This is on my iPhone 5s, which has bad low-light performance, as you can imagine. So I wanted to see if I could take a video (without the phone moving), and then merge all the frames in the video into a photo.

Either Mac or iPhone apps will do. I'm looking for something that's not command-line, is cheap ($10 at most) and easy to use -- like one-click software, rather than software that requires me to try many things and see what works well.

4

I'm not sure that this will work well due to how video works. Video is initially MUCH lower quality than photos. 1080p video is only around 2 megapixels of image quality. Further, cellphone video does not record each individual frame of video, but rather records the changes from one image to the next and only records a couple of photos per second or less on average. This means that the actual noise information is not going to be well preserved and artifacts from compression will result in adding noise to the attempt to form an image.

If you had a camera that supports capturing all I frame video, then that would potentially be useable, but anything that uses IBP frame video is not going to be ideally suited for this purpose.

  • 1
    Exactly what I was thinking. In fact, since he is taking a video of non moving scene, there is relatively no or little change between frames, so the information he wants to stack simply doesn't exist. You will pretty much just have a couple of low quality images. – Robin Jan 14 '14 at 17:33
  • Isn't the B and P frame thing just a question of compression? The camera is capturing a frame every 1/30 seconds; they are just encoded in a lossy format. BTW, I was thinking as to whether the iPhone 5s supports full sensor readout. If so, the 2 megapixel resolution shouldn't be a problem. – Vaddadi Kartick Jan 14 '14 at 17:52
  • @KartickVaddadi - B and P frames contain only changes from the previous frame, not full frame information. Thus, they are not going to have a sensor noise pattern as that will all be lost in the compression. They are going to instead have artifacts based on the attempt to relate each frame to the previous frames. The camera does capture 30 frames per second, but it only stores a small fraction of that information. (Typically only 1 or 2 actual full frames per second with the rest being changes only.) – AJ Henderson Jan 14 '14 at 18:20
  • At what point does the noise become "loud" enough that the compression algorithm sees it as frame-to-frame changes? – Michael C Jan 14 '14 at 20:41
  • 1
    @MichaelClark - that will depend entirely on the data rates involved. It would be hard to measure, but in general, video compression is exceptionally lossy compared to still photo compression since it is already relying on the high speed playback of frames to cause our eye to reconstruct a high quality image. – AJ Henderson Jan 14 '14 at 20:43
5

You could try using Registax. Generally, Registax is used for astrophotography, particularly of the planets in our solar system. Registax uses the concept of superresolution to stack hundreds or thousands of frames, discard the worst, keep the best, then interpolate the information from all of those frames in such a way that it enhances detail and resolution.

While it is not specifically designed to do what you want, I see no reason why it couldn't be used for the purpose you are describing.

As an alternative, you could also use the median merge approach. Instead of using video and stacking to "photograph" low light scenes, just take about 10 still frames of the same scene (assuming no motion of anything within it) with the iPhone on a tripod of some kind. With Photoshop, you create a layer out of each of these still frames, select them all and convert to a smart object. Once you have a smart object, you set the stacking mode to median, which will GREATLY reduce the amount of noise in the scene without affecting detail.

Both solutions should work. The registax approach would be a little less manual work, however it might not produce as good a result as the manual stacking approach.

  • Is 10 still frames the point of diminishing returns? That is, 20 aren't going to be noticeably better. BTW, I don't have Photoshop, but I do have Lightroom, for what it's worth. Registax seems to be Windows-only :( As I wrote in my question, I'm looking for Mac or iPhone software. Can Photomatix do median (or average, for that matter) merging? – Vaddadi Kartick Jan 14 '14 at 17:55
  • You have to be careful how many frames you stack with median method, because stack too many, and you could start losing color fidelity as well. This is more a concern with higher ISO settings or shorter exposures. I usually use this technique with astrophotography as well...however even at high ISO, with astro, you have an even weaker signal than with a still shot at 3200 or 6400. You might be able to stack 20 without color fidelity loss (and the results would probably be mind blowing), just keep an eye on things other than just noise levels. – jrista Jan 14 '14 at 20:42
  • I am not sure what Photomatix can do these days. I stopped using that years ago. For Mac, the best tool is really probably Photoshop. You might try looking into other astrophotography stacking programs, see if any of the ones that support mac have a video stacking or "drizzle" feature. – jrista Jan 14 '14 at 20:43
4

I know that you asked for a non-command-line solution, but I think that this one is simple enough to post. Moreover, I noticed the "no CLI" only after I wrote it, so... maybe could be useful for someone else.

There is an example doing exactly that (although for a different objective --- long exposure without ND filters) that I think it is what you need in this blog entry by Patrick David. Look at the end where he says "Something neat".

He uses the open source programs Hugin and ImageMagick to achieve the treat. In my Linux machine, without aligning, it's just one command:

convert test.mpg -evaluate-sequence mean -alpha off OUT.tif

(to align, it's a bit more complex, but still you can do it in a script easily).

Also interesting could be the use of a median filtering, simply changing mean to median above.

0

"Photoacute" is software that's designed to do frame-averaging and interpolation for noise reduction and super-resolution work from sequential still photos. I ran it years ago on a Mac, with good results, but haven't upgraded to the new version.

The iPhone 5s isn't on it's supported cameras list, but that doesn't matter for reducing noise by merging images. Sadly the 'lite' (non-raw, non-16bit) version is over-budget at $30, but perhaps some of its other tricks might be useful as well.

The website is here, but be warned that it's not the most sensibly laid out: http://www.photoacute.com/studio/index.html

0

There is an iphone App called Cortex Cam that takes 10 second video and converts it to a reasonably noise free photo. I have got good results with it sometimes.

  • Yes, thanks, that's one of my favorite apps, but while it reduces noise, the noise is still sometimes too high. I'm looking to create photos that my point-and-shoot could with 15-second exposures at ISO 80. Really clear, no noise, bright colors... – Vaddadi Kartick Jan 15 '14 at 3:55

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.