Band picture last night

Took my camera to see a really nice local Latin-American band last night -


Singer is Indira Roman.
EOS 6D, Samyang 135 f2, 1/125th, f2, iso2500. Processed in RT and name added in Gimp.

2 Likes

@RawConvert how well does the autofocus perform on that lens with moving subjects in low light? i.e. if you autofocus on a moving musician’s eyes, what chance that the photo will be sharp on the eyes?

Morgan, like most (I think) of the Canon-fit Samyangs, the 135 is a totally manual lens. So no autofocus in any light! I focused it in magnified Live View, then relied on the performer not moving around too much. I experimented some time back photographing cars going by with a Canon 50mm in “AI Servo” mode, and it worked quite well, but mainly, I seem to take static subjects. The 6D does indeed have face detection, but I doubt it locks onto eyes like some of the latest whizzo cameras.

1 Like

Another one just announced, see the Eye AF video part way down…

1 Like

this looked very cool (not that I will ever buy a 2k camera, but still) until I saw this… (zoom in on the girl’s face)

Eye AF seems like a difficult problem to solve, esp. in low light conditions. I have played with eye detection code in the past (not mine, obviously) and it tends to yield many false positives or true negatives.

If you ask the band members really nicely, maybe they will freeze enact for you :rofl:.

Okay, I did, and it’s not that sharp. But isn’t a lens starting from f4 the wrong glass for the job? So I guess you are referring to something different. Could you elaborate a bit?

How true. There’s better value for money around (IMHO).

Uh uh, I realized I linked the wrong picture, sorry! This is what I was talking about (1:1 crop):

But to be honest, dpreview has just updated their description writing that this can happen also with other cameras with on-sensor PDAF pixels. Which makes me want to try to reproduce on my a6000…

There are vertical stripes:
6941227258

@agriggio Thanks! Now it’s obvious!

@Morgan_Hardwood You even made it blatant. :grin:

…and, sadly, I could do that very easily :frowning:
DSC09882.ARW (23.7 MB)

Here’s how it looks like (especially visible on the red shirt of my daughter’s dear Pippi :slight_smile:

So, I started looking in my favourite “bag of tricks” (RT, that is), and I found the “line denoise” tool:

Definitely better, but some fine detail is lost. The challenge now is how to do the same without losing detail. I want to look into this, but suggestions are welcome!

1 Like

In the remote case that someone besides me is interested :slight_smile: , here’s the solution I came up with:
55

Result:

2 Likes

If the lines are regular, perhaps FFT pattern removal could be employed. Might not work as well or simply as your solution…

If I understand the code correctly (and that’s a big if, considering how little I know about signal processing), I think that’s similar to what the line noise filter does already: it performs a DCT and then uses a statistical method to distinguish noise from signal. I’m playing a little bit with the FFT support in the G’MIC plugin, and the results I get are indeed somehow similar… somehow :stuck_out_tongue:

Likely. I guess the advantage of having access to the spectrum directly is that if you know what you are doing you would have much more flexibility over what you could do with the image.

sure, if you are comfortable editing fft images in GIMP, I think you can do better for sure. the filter is meant to be a 1-click (more or less) solution for us mortals :slight_smile:
it’s definitely not perfect, but in my tests good enough to turn this from a “deal breaker” to an “annoying glitch” (again, all hypothetically speaking as I will never buy an a7iii :slight_smile:

1 Like

Definitely; hence the strong if. Good to know what the line noise filter actually does :slight_smile:.