August 02, 2014, 02:12:16 AM

Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Messages - Marsu42

Pages: 1 ... 3 4 [5] 6 7 ... 291
Photography Technique / Re: 85 vs 135 for portraits
« on: July 09, 2014, 11:39:57 AM »
The difference between 100 and 300 was minimal to my eye.
Other factors become more important, unless you are really after that perfect in your eyes look.

One factor I'd also like to mention: Having more distance can be beneficial for portraits since the subjects feel less bothered by "lens in your face" photography and you can get very natural shots when they don't realize how tight the framing actually is with 300mm. It's a pita to focus though.

Photography Technique / Re: UWA odd angle postprocessing howto
« on: July 09, 2014, 11:16:54 AM »
I find wildlife more fun, too, particularly because it is so challenging, which makes it that much more rewarding when you get the shot you want.

Problem is: If people don't realize how complicated this is, it can also be kind of frustrating because they're used to studio-type shots. That's why I seldom shoot insects nowadays, killing/capturing them simply produces better shots than crawling through the woods. I rather go for animals or environments that are visibly "outdoors" and non-staged.

My plans now that I have the 1D X with its Ethernet interface are to take UWA shots of the more dangerous and skittish wildlife from a distance.

Um, with what - robotic gear to move and aim the camera? My 6d has wifi control, but lv focus is slow (get a 70d for that), and you really really have to setup the scene like "squirrel grabs food" which I didn't come around to yet. I'm looking forward to see some results with this :-)

What type of horses are those?  I know you've posted them before and I have meant to ask you.  I think they're beautiful and interesting subjects and I'm surprised they are so tame.

Well, tame to me, I know them for 2 years and it wasn't always the case - I'm now rather fluent at speaking horse and know when to step aside. Imho still much safer than driving a car, or a bicycle surrounded by cars for that matter.

They are horses of the "Konik" type, it's a project to back-breed the extinct European "Tarpan" wild horses which as usual where hunted down until none was left. For me "wild horse" doesn't mean "bad temper all the time" but "do what they want to do and nothing else" which can include being curious and playful. They're always happy to have another friendly set of eyes and ears around to watch out for predators :-)

They are often used in landscape projects in Europe (PL, D, NL) and are basically not cared for at all, they are outside all the time w/o any man-built shelter and have to look for food themselves. This means they've got an actual archaic herd structure and act like most horses would act if people wouldn't keep trying to ride them.

You really gain respect for them when they are resistant to about every weather or disease/injury and survive in the hard winter living by digging up roots when you as a human wouldn't last for a day. In comparison to their senses, I feel blind, deaf and slow as a slug:

For the shots with no reference lines that you can't seem to get right, I'd try the following techniques [...] See which one looks best from a slight distance and that's the one to fine tune or use as is.

This certainly good advice, and I'll try to remind me of this - just like dof, "angle" seems to be very dependent on print/export size, what looks ok at small size or very near looks odd when being at a distance.

For these, *everything* has to be spot-on, not just the technical aspect but also the emotion or expression of the scene. But it's really rewarding to get this done as you don't find these shots anywhere. They even come with real non-digital rainbows :-)

Photography Technique / Re: 85 vs 135 for portraits
« on: July 09, 2014, 11:10:02 AM »
There you go:

Great link, thanks - you actually learn things around here :-) ... and looking at these I keep on feeling fine with my "poor man's gear" 70-300L/4-5.6 and 100L/2.8 combination.

EOS Bodies / Re: Which is better for high ISO, 6D or 5D Mk III?
« on: July 09, 2014, 11:07:38 AM »
Which is why I prefer fast primes over f4 or f2.8 zooms. I gain a far lower iso value or a higher shutter speed.

Well, but lower than "100" is "screw on nd filter" and the iq of 1600 on the 6d is so good I seldom feel the necessity for a really fast prime. You can post-process push a bad exposure, but cropping from 20mp (too wide prime) is tricky, and adding image data in post (too long prime) is impossible.

And I already raved about how damn difficult it is to af with the 6d in enough threads, and even with a better af system the dof of f4 on longer focal lengths is already very thin @100% crop. Not that I'd dismiss fast primes as you get thin dof @small export/print sizes, but I guess I'm the "zoom flexibility" type of photog.

Photography Technique / Re: 85 vs 135 for portraits
« on: July 09, 2014, 08:31:46 AM »
Is 135 not versatile enough for half body shots?

Well, you might want to step back a little ... but no problem if there's space. If.

Imho the main difference is compression (I'm using my 70-300L for portraits so I know which zoom setting I like) - 85mm is rather wide on ff so if you're getting closer to the subject it doesn't look nice anymore, big nose and such. Having the option for very thin dof 135mm shots will separate you more from the crowd. If you manage to focus where you want. Again, if.

So for example a 300 mm lens won't be more 'flattering' than a 135 but you will get considerably more subject isolation, but you'd need much more space and more light.

Ugh? Is this your personal experience or are you having some charts available to support this theory? Not that I'd dispute what you say, but right now my impression would be that 300mm *does* have significantly more compression than 135mm even to the extend that 300mm looks to "flat" for my taste.

EOS Bodies / Re: Which is better for high ISO, 6D or 5D Mk III?
« on: July 09, 2014, 08:23:53 AM »
I have a 6D, which I've been using in low light situations. In recent months I've found myself frequently exploring the limits I can go and still come up with what I consider successful photos. I've very happy with the results I've been getting up to ISO6400 or so.

After downsizing 22mp 5d3 -> 20mp 6d and adjusting for slight sharpness differences it's a wash, this is the same sensor generation. However the 6d has significantly less banding, i.e. higher dynamic range which *does* make a difference if your shots require it - it even tops 1dx on low iso. You can boost the dr on 6d even higher with Magic Lantern, nearly up to 15ev.

Having said that, no nice noise performance makes up for dodgy metering or getting out of focus shots with the ancient 6d af unless you're mf'ing (like with ML's focus peaking).

Even what I've shot at 12800 pleasantly surprises me.

ISO 12800 is just a digital push (on 6d, actually even 6400 is) - no need to use it unless you shoot jpeg or do video, otherwise underexpose a lower iso and keep the higher dynamic range of the lower setting.

The main positive about the 6D is the centre point is rated to -3ev whereas the 5DMKIII is -2EV.

Well, yes, it's "nice to have" esp. with slower lenses like f4+. *But* af'ing @-3lv is so slow that you could also just switch to live view and contrast af or mf with a boosted lv (ML again) which works far below -3lv.

Photography Technique / Re: UWA odd angle postprocessing howto
« on: July 09, 2014, 07:55:22 AM »
To me, only the last one might be a little off to the ccw direction, based on the trees, although the horizon looks about right at first glance.

That's exactly what I'm talking about: Something always looks wrong, it's about *minimizing* the odd parts... for example the middle shot with the water has a completely "wrong" rotation vs. gravitation direction.

I would like to come up with some guidelines to speed up the trial & error process, but failed so far. Btw the first shots were really some random holiday shots, they just happened to have no right angles in them so they're ok for demonstration purposes.

One thing I can say is that vertical lines seem to be more important than horizontal lines, at least my eye is used to sloped horizons but not so much to skewed trees. It becomes tricky when the tress actually *aren't* vertical, in this case you need to shoot some space around to let the viewer see it and get some bearings or it tends to look wrong.

Reviews / Re: Is Canon 5d mk 1 still a good camera?
« on: July 09, 2014, 04:12:01 AM »
That's what is happening to me: I can't convince myself that I need a new Full Frame body, so I'm keeping my 5Dc plus my EOS M when I need to shoot videos or to travel light.

The 5d3 has full res raw video capability (with Magic Lantern) so that's a decisive feature added, of course apart from everything else ML offers as it isn't available on the 5dc.

Otherwise only upgrading a body after knowing what exactly the required improvement is imho is the best advice. Much more sense putting $$$ into lenses and esp. lighting gear (1x 600rt = €500 ...). It's easy to test-drive a new camera nowadays either via rental or in a shop, if it doesn't immediately deliver the "that's worth a couple of thousand $$$" impression the money might be better saved.

Photography Technique / Re: UWA odd angle postprocessing howto
« on: July 09, 2014, 01:26:42 AM »
These samples are all pleasant to the eye, so I think you're on the right track.

Thanks :-) ... the thing is that these are samples of shots I had to turn in 0.1 degree steps since I wasn't able to find a good rotation right away. The important point I've learned is that there doesn't seem to be a magic bullet solution for these.

Your advice is still valid though, esp. concerning the vanishing point - it's just that I seldom do architecture these days, "been there, done that" :-) and I now know how to nd-filter, focus-stack and bracket for good results. Tedious setup though. Crawling through the mud shooting wildlife is much more fun :-> and I find it esp. challenging to get different results from a "my mobile phone also has this focal length" type shots.

Photography Technique / Re: UWA odd angle postprocessing howto
« on: July 08, 2014, 04:33:59 PM »
I don't really have a solution for you, but I can totally relate. I have a Sigma 8-16, which I really love, but I also always struggle nailing the right tuning of the image.

Good to hear I'm not the only one, I often fear I'm sounding completely clueless when posting threads like this :-p esp. since I know I'm "horizontally challenged" and need a grid inside the vf.

Taking into account perspective, and looking at the banner on the ceiling, it looks like it needs to be rotated CCW a little.

Thanks for reminding me, I even vaguely remember even learning this in school and drawing perspective wireframes in art class, before computers became popular that is :->

Shoot from a tripod and level the camera

To me, this only works if it's a "take all in" type shot and the eye has references to guess where "up" is - but with many closer up uwa shots this looks wrong esp. if there are skewed lines in the picture that aren't aligned.

Try to compose your shot using one or more perspective vanishing points

I do this with architecture, and with excellent post-processing tools like dxo viewpoint (or ps) you usually can get some pleasing result even if you failed on the scene - though at a loss of resolution w/o a ts lens.

However, I'm currently mostly using my uwa for wildlife to get a "close up" effect. In these cases, I have to "balance" a non-level background against a non-level foreground in 0.1 degree steps until it looks least strange because w/o a clean background it's very hard to get an impression about the actual perspective/vanishing point or sometimes there's no horizon at all:

Photography Technique / UWA odd angle postprocessing howto
« on: July 08, 2014, 07:51:38 AM »
Since I've got a ff and my 17-40L I'm more and more leaning to the wide end because the results look interesting to my eye.

Problem: I often have difficulties finding a horizon or post-processing rotate angle that looks ok-ish. This turns out to be most difficult if it's an odd-angle shot and there are no orientation lines or the lines aren't right angle themselves (like skew trees in the background or a sloped/non-even horizon on a hill).

What I'm doing is right now is subjective trial and error ("nah, another 0.1 degrees clockwise") but whenever I look at it again it still seems to be somehow odd. Most annoying is that a "gravity down" approach with the camera's sensor also doesn't always look correct. To me :-p

Question: Do you know or have developed any guidelines on how to find the "best" post-processing angle? Do you often do tse-like perspective correction on these shots? Or do you simply accept the fact that these shots never look right unless you're there and your body know where the gravity is?

Um, I'm rather confused at why the ML guys couldn't figure out the temp/tint model. That's actually the same as the L*a*b* color space. It's just blue/yellow and magenta/green opponent-process color axes, which are the same opponent-process color axes human vision uses (i.e. we cannot simultaneously see blue and yellow at the same spatial location...same goes for magenta/red and green...this is due to the way the cones of our eyes respond to light via a tristimulus factor.) This stuff is pretty thoroughly researched theory (decades old theory)...if the ML guys want to figure it out, they should probably look into color theory, especially the work done by the CIE.

Well, alex did add two wb autotdetection algoritms, "greymed" and "greymax". As far as I understand it the latter to maximize the grey areas by adjusting the wb, the former is a "traditional algorithm" as alex puts it. Both often work ok, but not in all cases and currently completely fails on underexposed shots, the major problem seems to be the tint value. No idea where he got the idea from, I have no insight into this at all.

Also, as far as I know, when I went searching, Canon provides fully functional source code for how they do...everything. It is not well documented, but if your a coder, and you can get your hands on source code, you should be able to figure out what it's doing.

I guess the ML people are not used to receiving any help from Canon, source code or otherwise. Canon's awb has got some "intelligence" built in that corrects their computed values for the specific scene type they figure is being shot.

You're saying Canon's awb algorithm can be found in the Canon Digital Camera SDK for everybody to see? Isn't this rather a closed source thing seeing how much Nikon struggles with getting the "correct" tint values?

Reviews / Re: Is Canon 5d mk 1 still a good camera?
« on: June 21, 2014, 03:03:57 AM »
Having said that, I will probably prefer a used 5DII for $1000 than a 6D for $1300. Unfortunately, all the used 5DIIs seem to be available for not less than $1300, so go figure!

That doesn't mean anything, subjective demand is no indicator for technical superiority. The dslr crowd in general is rather conservative and old-school, also look at the prices for older L lenses that are rather mediocre by today's standards.

I put a lot of time deciding between 6d and 5d2, and the 5d2 is more "pro" as in full mag body, higher shutter rating, 1/8000s shutter, 1/200 x-sync and cf cards, but otherwise the 6d is the more modern and superior camera:

EOS Bodies / Re: New Sensor Tech in EOS 7D Mark II [CR2]
« on: June 20, 2014, 05:37:10 AM »
It isn't the same as what ML is doing. They are either reading alternate lines of the sensor at two different ISOs, then combining them at HALF THE RESOLUTION, or they are doing two full reads of the sensor.

ML dual_iso is interlacing the frame with another iso at each other line.

What's really amazing is that the postprocessing program ("cr2hdr") does such a great job at reconstruction because you'd think you loose a lot of resolution. But since in the real world few people are shooting test charts or scenes that contain such fine detail to cover only one horizontal scanline, ML knows what "probably would have been there" if it wouldn't have been clipped.

It's only in completely black or white areas that are only constructed from half of the scanlines you see the loss of resolution, but after shooting ~3000 dual_iso shots I can say you have to really look closely and the benefits outweigh the drawbacks by far.

Looking back, a huge amount of my vanilla daytime shots on 60d w/o dual_iso or brackting have clipped sky somewhere, now with  (higher dr) + dual iso you see it's actually still often blue even if you think it's white. It would be great if Canon would do this in-camera and output 16bit raw files like ML does.

Yeah, it's kind of a mess. It really isn't any better for other manufacturers. Some of them have even more radical generational changes in their metadata than Canon does.

It would be interesting to know if the Canon guys are as confused as the rest of the world is by now, or if they've got top-notch internal docs and samples that make everything easy to do.

For example I know the Magic Lantern devs recently failed to figure out Canon's awb algorithm - it's a complete mystery what all these color channel tags exactly mean and how they end up in a temperature and tint value (they need it for the dual_iso module). Who knows how different that is between camera models. First they tried to compute the awb from the tags which did very seldom work, now they compute it from the ground up looking at the pixels which often also doesn't work :-\

Pages: 1 ... 3 4 [5] 6 7 ... 291