ProfHankD

ProfHankD

Lives in United States Lexington, United States
Works as a Professor
Has a website at http://aggregate.org/hankd/
Joined on Mar 27, 2008
About me:

Plan: to change the way people think about and use cameras by taking advantage of cameras as computing systems; engineering camera systems to provide new abilities and improved quality.

Comments

Total: 1250, showing: 1 – 20
« First‹ Previous12345Next ›Last »

It's interesting that as still and video cameras converge, lenses for them are quickly diverging. It seems like the last year or two just about every lens maker has come out with cinema-oriented versions. Are they really selling that well or is it simply that higher pricing (and higher markup) are tolerated better in that market?

Link | Posted on Apr 23, 2017 at 22:58 UTC as 3rd comment
On article Samyang Lens Station USB dock spotted in the wild (21 comments in total)
In reply to:

PhotoKhan: The simple fact that a brand has to offer a "Lens Station" for "firmware updates and customization" is an important red flag in what relates to buying 3rd party lenses for proprietary mounts.

Take Sigma's case, for instance.

In spite of also having said "dock" and having vastly improved the accuracy of the AF on their EF lenses working on Canon bodies, they simply seem unable to sort out the consistency side of the equation.

You might be doing fine with a Sigma lens on a Canon body, beautifully accurate AF photos pumping out of the camera and then, all of a sudden and without reasonable explanation, the lens decides to focus on something else for a particular image or series of images.

A lot of the point in this comes down to the camera manufacturers not having open specs for their lens interfaces. If they were, you'd be able to do this stuff via the camera... as many cameras can do firmware updates for native lenses.

Link | Posted on Apr 23, 2017 at 09:16 UTC

Hmm. It's a Canon/Fuji party with a rather narrow range of lens types....

Link | Posted on Apr 22, 2017 at 12:33 UTC as 16th comment
On article Hands-on with the Panasonic Leica 8-18mm F2.8-4 (155 comments in total)
In reply to:

Quest21: Come on Panasonic it's time to make a camera with iso50 for low noise. With this lens it would be great for landscape or cityscape...

Not how it works. Low ISO with a high-sensitivity sensor means you need high well capacity... and small sensor pixels don't get that. You'd just clip dynamic range, which isn't a strength for MFT now.

Link | Posted on Apr 19, 2017 at 10:22 UTC
On article Light's L16 camera is in final stages of testing (300 comments in total)
In reply to:

orion1983: ...and what is the size of a single 23MP sensor? I could not find this info....

Argh! Forgot to mention the high-res images are 5163x3872, i.e., not quite 20MP. Scaling these images down to 12MP and back up to 20MP actually improves the IQ by smoothing some JPEG compression artifacts (identifiable by their 8x8 DCT pattern); the level of scene detail seems to be about what should be present in a single good 12MP capture, and the lighting should have allowed good captures.

Just to be clear, I am a very strong proponent of computational photography methods being used in cameras -- but I strongly oppose technologies being oversold. For example, overselling abilities is why the state of the art in consumer-level 3D printing has been nearly stagnant for several years: people made claims beyond what the tech could do, so there has been relatively little excitement/investment in making the tech actually able to do what was claimed.

Link | Posted on Apr 16, 2017 at 13:11 UTC
On article Light's L16 camera is in final stages of testing (300 comments in total)
In reply to:

orion1983: ...and what is the size of a single 23MP sensor? I could not find this info....

falconeyes: just to close this discussion, with the pseudo-random camera placement, the odds that they can't find a pair of non-occluded cameras for each scene feature are very low, so synthetic bokeh should be reliable if not computationally cheap. Enhancing overall IQ and resolution by as much as they claim is MUCH harder, and might even be impossible for most scenes, especially at closer focus distances.

They only give 3 high-res images on their site, two of which are more than sufficiently distant so that dumb stacking (without any significant alignment) will work. The middle sample image is a portrait that is close enough to have issues, and it's pretty good, but does have some issues -- e.g., there's a strange shadowing (ringing?) around the back edges of the hair rather than normal blurring of the edge. That defect is consistent with making a failed attempt to combine edge points that didn't really align due to occlusion.

Link | Posted on Apr 16, 2017 at 12:35 UTC
On article Light's L16 camera is in final stages of testing (300 comments in total)
In reply to:

orion1983: ...and what is the size of a single 23MP sensor? I could not find this info....

falconeyes: Hmm. If they think just simple alignment and stacking will give them the improved IQ they claim, I think they're wrong unless they don't let you focus closer than about 30*max distance between lenses (which would be several meters)... that's roughly where occlusions start to become less significant.

There is no such thing as "an array of lenses with no gaps" at the level of the bokeh (especially if those lenses have variable apertures). Bokeh are extremely sensitive to tiny flaws, making it very difficult to produce high-quality apodizing filters; e.g., the STF lenses make an optical flat from a plano-concave smoked glass and a plano-convex clear glass with the same refractive/dispersive properties.

Yes, noise and featureless areas are problems for stereo matching, but because neither of those have significant image content, you can't really see the screw-ups very much. ;-)

Link | Posted on Apr 16, 2017 at 01:22 UTC
On article Light's L16 camera is in final stages of testing (300 comments in total)
In reply to:

orion1983: ...and what is the size of a single 23MP sensor? I could not find this info....

falconeyes: parallax normally just requires stereo matching to get the depth and then computationally render simulated bokeh (OOF PSF) -- you compute the diameter from the distance between corresponding points and then paint an OOF PSF of that size. The catch is that occlusions produce stereo matching problems that have no solution. Any dense regular pattern that has fewer lenses than there are pixels in the OOF PSF diameter would still require stereo matching to synthesize correct bokeh, so that really doesn't help much, and sparseness is probably a good thing.

The problem is that I have seen nothing from Light that suggests to me that they have any useful insights into the needed computation... let's hope they do and just haven't said anything about it yet. ;-)

Link | Posted on Apr 15, 2017 at 20:41 UTC
On article Light's L16 camera is in final stages of testing (300 comments in total)
In reply to:

orion1983: ...and what is the size of a single 23MP sensor? I could not find this info....

Melchiorum, stacking from different camera positions is problematic because of issues like occlusion. The closer scene objects are, the more the distinct viewpoints result in missing/inconsistent data. Of course, the IQ should never get worse than a single one of the component cameras would produce... and lots of people are happy with cell phone camera images having fewer than 13MP, especially in good lighting conditions.

Link | Posted on Apr 15, 2017 at 13:20 UTC

Hmm. Standard umbrella hats are pretty unstable in a high wind -- I have enough trouble keeping a hat on my head. This device looks like it securely anchors to your body via a padded backpack-like strap system and seems to have a better wing shape. I wonder if one could get enough lift to get airborne wearing one of these? ;-)

Link | Posted on Apr 15, 2017 at 05:38 UTC as 75th comment | 2 replies
In reply to:

stevo23: SCOTCH whiskey. I'm just sayin'

Now if it were a bottle of the real deal Pappy Van Winkle...watch out

It's true. Top-quality Scotch Whisky is literally what they make reusing the barrels that are no longer fit to make Kentucky Bourbon (Bourbon is always matured in fresh casks to extract the maximum flavor). Nice to celebrate photographers, but speaking as a Kentucky resident, the "kit" packaging seems a bit desperate.... ;-)

Link | Posted on Apr 14, 2017 at 01:15 UTC
On article Panasonic Lumix DC-GH5 Review (1152 comments in total)
In reply to:

snapa: It still amazes me how a 'm4/3' camera can cost $2,000 (without any lens), and be considered such a great camera. Once you start adding Pro lenses, then think about how well the IQ is at higher ISO levels with still pictures, it does not make any sense to me.

Maybe if you are looking for a very good 'video camera', it would be a good solution to your problem if you are looking to get very good videos. Still, to me, a m4/3 sensor camera for $4-5,000 (with lenses) that take just OK still pictures seems like quite a bit of money to me IMO, considering its competition with larger senor cameras.

BTW, how may professional videographers seriously look to buy a m4/3 cameras for taking serious video?

Astrotripper: all your complaints are on video, and I think they're a bit overstated; heck, the GH5 isn't even coming with log video support by default (and it has obviously poorer DR). In my opinion, the A6500 beats every micro4/3 body for stills by a much larger margin than it loses by on video. I also think $1400 is a lot less than $2000... and if it isn't, then an A7SII, A7RII, or A99II isn't unreasonable either (the GH5 is 42% more than A6500, and A7RII is 45% more than GH5).

In sum, the GH5 is an impressive enough camera, but $2K is too much for what is ultimately a big, video-centric, body wrapped around a small sensor with an aspect ratio that for video discards at least another 25% of the sensor due to choice of the 4/3 sensor aspect ratio. Your opinion may, and is welcome to, vary. ;-)

Link | Posted on Apr 12, 2017 at 14:54 UTC
On article Panasonic Lumix DC-GH5 Review (1152 comments in total)
In reply to:

snapa: It still amazes me how a 'm4/3' camera can cost $2,000 (without any lens), and be considered such a great camera. Once you start adding Pro lenses, then think about how well the IQ is at higher ISO levels with still pictures, it does not make any sense to me.

Maybe if you are looking for a very good 'video camera', it would be a good solution to your problem if you are looking to get very good videos. Still, to me, a m4/3 sensor camera for $4-5,000 (with lenses) that take just OK still pictures seems like quite a bit of money to me IMO, considering its competition with larger senor cameras.

BTW, how may professional videographers seriously look to buy a m4/3 cameras for taking serious video?

This camera benefits greatly from its DPReview classification, which doesn't pit it directly against, for example, the MUCH CHEAPER Sony A6500. This seems like a very good camera, and excellent for its sensor size, but it trails pretty significantly as a $2K body-only stills camera. As a video camera with rolling shutter issues, $2K doesn't really class it as a bargain either, although it's more competitive. I don't think "gold" rating is justified at this price point (especially when the A6500 is "silver"), but micro4/3 definitely has a dedicated following willing to pay more for high-featured big cameras with little sensors....

Link | Posted on Apr 12, 2017 at 03:08 UTC
In reply to:

Internet Enzyme: Sees Headline.

5K video oh man that sounds real fancy but I'm positive it has a horrible bitrate since all of these 360 degree and drone cameras have horrendously low bit rates.

Reads article.
"Videos are recorded using the H.264 codec with a bit rate of 60Mbps,...".

Come on people. When will they know that high resolution doesn't matter if your bitrate is so low.

The real quality issue is stitch errors. This seems to do very well, with only one stitch error obvious to me in the sample video. The rule of thumb is no closer than 30X the difference in camera reference points, which here I'd guess means about 5-6 feet... so no 360-degree macros unless they are carefully aligned within one camera's view. It is kind of odd they don't support a larger battery/storage unit for longer runtimes and higher bitrates....

Link | Posted on Apr 11, 2017 at 01:06 UTC
In reply to:

ProfHankD: In other words, Sigma has discovered that they can spit out a color-interpolated "uncompressed" TIFF file like many cameras did 15 years ago. DNG is just one of many variants of TIFF, and all using the DNG marking here buys is the ability to use 12 bits per pixel color channel, while uncompressed TIFFs normally were 8 bit (or 16 bit).

The code in dcraw for Foveon interpolation carries some restrictions that are problematic for tools built using dcraw code (which is nearly all software that can process raws). I think the better answer for Sigma would be to distribute raw decode source code without restrictions....

Mgrum, the code doesn't look all that clever to me. However, there's a good chance the issue isn't cleverness nor giving competitors a leg up, but rather a bit of legal ambiguity about who gets to open it. Quite often, there's code that isn't owned outright, but is purchased as IP with restrictions; it also wouldn't surprise me if legal oddness came about when Foveon became part of Sigma. Anyway, it's really not helping Sigma/Foveon to have it closed.

Link | Posted on Apr 10, 2017 at 15:07 UTC
In reply to:

ProfHankD: In other words, Sigma has discovered that they can spit out a color-interpolated "uncompressed" TIFF file like many cameras did 15 years ago. DNG is just one of many variants of TIFF, and all using the DNG marking here buys is the ability to use 12 bits per pixel color channel, while uncompressed TIFFs normally were 8 bit (or 16 bit).

The code in dcraw for Foveon interpolation carries some restrictions that are problematic for tools built using dcraw code (which is nearly all software that can process raws). I think the better answer for Sigma would be to distribute raw decode source code without restrictions....

From wikipedia: "Tagged Image File Format, abbreviated TIFF or TIF."

DPReview tends to think in terms of Adobe raw processing being the standard. Adobe does what Adobe does, and since they own DNG, it shouldn't be shocking that they handle it well. Adobe couldn't just lift the Foveon code from dcraw (as they essentially did for other raw formats) because of the Sigma use restrictions, but using dcraw has long worked fine for Foveon interpolation, so image editing tools literally using dcraw as a plug-in have never had a problem.

Dave Coffin is really good about adding support for new raw formats to dcraw, so I'm sure pixel shift stuff will be in there soon... actually: https://www.dpreview.com/forums/post/55912470

Link | Posted on Apr 10, 2017 at 02:04 UTC
On article Sigma sd Quattro H real world samples gallery (108 comments in total)
In reply to:

Gary Dean Mercer Clark: Here is a review that has sliders that compares the output of the Quattro H with the Canon 5Dr. This camera is deffinitely shooting above its pay grade. :) if you use Chrome browser and enable google translate, it will automatically translate this article. I'm pretty impressed with how the Sigma H images hold up to the 50MP Canon images. Excellent job Sigma!
https://www.focus-numerique.com/hybride/tests/sigma-sd-quattro-h-2977.html

Good review you pointed at there. Basically, this camera is a real outlier -- fundamentally very different from the Canons, Sonys, Fujis, etc. If what you want is monochrome resolution or freedom from moire, this is a winner -- and resolution/$ is great. If you want very accurate colors or low noise (especially at high ISOs) it's beaten by APS-C bodies costing 1/3 what it does. Arguably, the noise comparison with 24MP APS-C noise would be more fair if the Quattro H images were SINC-scaled to about 6MP and then SINC-scaled back to 24MP... then they look a lot more competitive in both noise and effective resolution.

Link | Posted on Apr 9, 2017 at 14:37 UTC
In reply to:

Tungsten Nordstein: 'Foveon sensors don't directly capture red, green and blue information'

Is this an accurate statement? One layer per RGB channel, surely means that they do directly capture red, green and blue information. See this diagram:

https://library.creativecow.net/articles/gondek_mike/Foveon/foveonchip.gif

Anyway, the fact that Sigma quattro cameras can produce RAW is good.

They are separate layers (photosites), but they aren't filtered very precisely.

Link | Posted on Apr 9, 2017 at 14:04 UTC
In reply to:

Tungsten Nordstein: 'Foveon sensors don't directly capture red, green and blue information'

Is this an accurate statement? One layer per RGB channel, surely means that they do directly capture red, green and blue information. See this diagram:

https://library.creativecow.net/articles/gondek_mike/Foveon/foveonchip.gif

Anyway, the fact that Sigma quattro cameras can produce RAW is good.

The top layer sees all colors, but light of different wavelengths statistically penetrates to different depths. Thus, you directly get very high quality monochrome data for every pixel site, but rather sloppy color sampling that requires significant computation to clean up. They are now doing that compute in camera to make a 12-bit uncompressed TIFF (which is marked as DNG).

Link | Posted on Apr 9, 2017 at 11:18 UTC

In other words, Sigma has discovered that they can spit out a color-interpolated "uncompressed" TIFF file like many cameras did 15 years ago. DNG is just one of many variants of TIFF, and all using the DNG marking here buys is the ability to use 12 bits per pixel color channel, while uncompressed TIFFs normally were 8 bit (or 16 bit).

The code in dcraw for Foveon interpolation carries some restrictions that are problematic for tools built using dcraw code (which is nearly all software that can process raws). I think the better answer for Sigma would be to distribute raw decode source code without restrictions....

Link | Posted on Apr 9, 2017 at 11:12 UTC as 70th comment | 4 replies
Total: 1250, showing: 1 – 20
« First‹ Previous12345Next ›Last »