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: 512, showing: 1 – 20
« First‹ Previous12345Next ›Last »
In reply to:

ProfHankD: Interesting. The Samsung NX1 raws look class leading (as they should be for a backside-illum APS-C sensor), but the Sony A6000 JPEGs are much nicer, especially at high ISOs (way less smeary). Well, I guess we know the next thing for Samsung to work on.... ;-)

Well, not to be inflamatory, but Canon's the only sensor maker that really sticks out in a bad way these days.... :-(

Direct link | Posted on Dec 23, 2014 at 14:18 UTC

Interesting. The Samsung NX1 raws look class leading (as they should be for a backside-illum APS-C sensor), but the Sony A6000 JPEGs are much nicer, especially at high ISOs (way less smeary). Well, I guess we know the next thing for Samsung to work on.... ;-)

Direct link | Posted on Dec 22, 2014 at 18:29 UTC as 28th comment | 3 replies
On BPG image format aims to replace JPEGs article (206 comments in total)
In reply to:

ProfHankD: Fabrice Bellard is the guy behind TCC (Tiny C Compiler), QEMU (a generic machine emulator and virtualizer), and FFMPEG. It's not easy to impress me with one's programming skills and knowledge, but Fabrice consistently awes me. I wasn't aware of the BPG work he's been doing, but I can say without doubt that it's worth taking seriously....

Incidentally, we all know good old JPEGs are riddled with artifacts. There are aspects of JPEG2000 that largely fix those problems, but standards that come before easily-adapted public implementations don't prosper. PNGs are not a very efficient way to encode natural scenes. In sum, an H.265 variant for stills with a public implementation, and potential to leverage video compression hardware, sounds worth a look. Storing big files might not be the big issue it once was, but transmitting ultra-high-res images is becoming a bigger issue.

Yes, we strongly disagree. A lot of JPEG "detail" is really artifacting due to the DCT compression of 8x8 blocks:

http://aggregate.org/DIT/ITDP/img20.png

I'm not saying BPG should replace all image formats, but it seems about as big an upgrade from JPEG as PNG is from GIF.

Direct link | Posted on Dec 16, 2014 at 11:54 UTC
On BPG image format aims to replace JPEGs article (206 comments in total)
In reply to:

ProfHankD: Fabrice Bellard is the guy behind TCC (Tiny C Compiler), QEMU (a generic machine emulator and virtualizer), and FFMPEG. It's not easy to impress me with one's programming skills and knowledge, but Fabrice consistently awes me. I wasn't aware of the BPG work he's been doing, but I can say without doubt that it's worth taking seriously....

Incidentally, we all know good old JPEGs are riddled with artifacts. There are aspects of JPEG2000 that largely fix those problems, but standards that come before easily-adapted public implementations don't prosper. PNGs are not a very efficient way to encode natural scenes. In sum, an H.265 variant for stills with a public implementation, and potential to leverage video compression hardware, sounds worth a look. Storing big files might not be the big issue it once was, but transmitting ultra-high-res images is becoming a bigger issue.

The JPEG2000 codings look pretty good to me, although this is obviously not lossless JPEG and we don't have access to all the tunable parameters for JPEG.

To me, BPG seems to have very natural-looking artifacting at very high compression rates, basically a loss of textural detail while still keeping edges cleaner than most compression methods. I agree that it isn't awesome for the highest quality, but to fill a WWW browser window on a 4K display with a viable image.... It seems to be an incremental improvement, mostly in how natural the artifacts look,

Direct link | Posted on Dec 16, 2014 at 03:31 UTC
On BPG image format aims to replace JPEGs article (206 comments in total)

Fabrice Bellard is the guy behind TCC (Tiny C Compiler), QEMU (a generic machine emulator and virtualizer), and FFMPEG. It's not easy to impress me with one's programming skills and knowledge, but Fabrice consistently awes me. I wasn't aware of the BPG work he's been doing, but I can say without doubt that it's worth taking seriously....

Incidentally, we all know good old JPEGs are riddled with artifacts. There are aspects of JPEG2000 that largely fix those problems, but standards that come before easily-adapted public implementations don't prosper. PNGs are not a very efficient way to encode natural scenes. In sum, an H.265 variant for stills with a public implementation, and potential to leverage video compression hardware, sounds worth a look. Storing big files might not be the big issue it once was, but transmitting ultra-high-res images is becoming a bigger issue.

Direct link | Posted on Dec 16, 2014 at 02:17 UTC as 51st comment | 5 replies
On Canon EOS 7D Mark II Review preview (1176 comments in total)
In reply to:

ProfHankD: What you call "ISO-Invariance" here has long been discussed in the forums as "ISO-less" behavior. It's a good thing to test and know about.

BTW, I had students do projects on it last year and this year in my Cameras as Computing Systems course, and have a research paper titled "ISO-less?" accepted to appear at the IS&T/SPIE Electronic Imaging conference February 10, 2015.

You're right, it's not just the sensor... although a heck of a lot of it is on that chip. Image capture system architecture isn't bad, but it's the whole system, all the way to making a JPEG.

BTW, I meant to ask: the line pattern noise is repeatable or varies from frame to frame for the same scene? There used to be a lot of that kind of dynamic noise when ADCs were not on the sensor chip (transmission line effects; wires as antennas), and repeatable patterns were also common due to fab process variations in pixels, but I've not seen so much of either on high-end cameras lately.

Direct link | Posted on Dec 12, 2014 at 13:12 UTC
On Canon EOS 7D Mark II Review preview (1176 comments in total)
In reply to:

ProfHankD: What you call "ISO-Invariance" here has long been discussed in the forums as "ISO-less" behavior. It's a good thing to test and know about.

BTW, I had students do projects on it last year and this year in my Cameras as Computing Systems course, and have a research paper titled "ISO-less?" accepted to appear at the IS&T/SPIE Electronic Imaging conference February 10, 2015.

Yeah, ISO-less isn't right for describing sensor behavior... but neither is ISO-invariant. Basically, you get different characteristics for different mixes of analog/digital "amplification." My paper ends up looking at how to take ISO out of the exposure formula, instead picking parameters based on scene analysis in the camera.

Direct link | Posted on Dec 12, 2014 at 10:40 UTC
On Canon EOS 7D Mark II Review preview (1176 comments in total)

What you call "ISO-Invariance" here has long been discussed in the forums as "ISO-less" behavior. It's a good thing to test and know about.

BTW, I had students do projects on it last year and this year in my Cameras as Computing Systems course, and have a research paper titled "ISO-less?" accepted to appear at the IS&T/SPIE Electronic Imaging conference February 10, 2015.

Direct link | Posted on Dec 12, 2014 at 01:49 UTC as 204th comment | 4 replies
On WaterWeight rethinks the sandbag approach to stability article (77 comments in total)
In reply to:

coronawithlime: Maybe I could use this to replace my hip flask :)

Or vice versa. Any weight on an appropriate cord can work, including a cord you simply step on.

Direct link | Posted on Dec 10, 2014 at 03:36 UTC
On Tamron 16-300mm Di II VC PZD real-world samples article (102 comments in total)
In reply to:

ProfHankD: Please test multi-mount-option APS-C lenses on sensors that are at least the standard APS-C size -- not Canon's slightly smaller ones. IQ issues are most likely near the edges, and using Canon's smaller sensor hides edges that would be visible for Nikon, Sony, and Pentax, whereas using any of the others would trivially allow the Canon-cropped edges to be ignored.

Given the proliferation of full frame bodies, it might even be appropriate to test using one of them, because many of us sometimes use APS-C lenses on FF bodies. Certainly, it will be appropriate to test ONLY on FF when we have FF sensors in the 50MP+ range (i.e., FF sensors with roughly the same APS-C crop resolution as the best APS-C sensors).

APS-C is "Advanced Photo System -- Classic" which is a standard film format of 25.1 × 16.7mm with a 3:2 aspect ratio. Even the aspect ratio of digital APS-C sensors varies a bit, but the standard APS-C film area is about 26% bigger than sensors in Canon EF-S bodies! Canon's interpretation is an extreme outlier; most APS-C sensors are only about 13% smaller than film APS-C. Put another way, Canon's sensors are about 11% smaller than most, and still 6% smaller than the few smaller-than-normal-for-Nikon outliers. That is more than enough difference to hide potentially problematic vignetting when testing only on Canon sensors.

Incidentally, there is some ambiguity about how sensor size should be measured. The pixels encoded in JPEGs are normally surrounded by pixels used only to avoid edge interpolation artifacts, and those are surrounded by masked pixels used as a dark reference. Perhaps common APS-C sensors actually are very close to the standard size counting all pixels?

Direct link | Posted on Dec 9, 2014 at 04:05 UTC
On Tamron 16-300mm Di II VC PZD real-world samples article (102 comments in total)

Please test multi-mount-option APS-C lenses on sensors that are at least the standard APS-C size -- not Canon's slightly smaller ones. IQ issues are most likely near the edges, and using Canon's smaller sensor hides edges that would be visible for Nikon, Sony, and Pentax, whereas using any of the others would trivially allow the Canon-cropped edges to be ignored.

Given the proliferation of full frame bodies, it might even be appropriate to test using one of them, because many of us sometimes use APS-C lenses on FF bodies. Certainly, it will be appropriate to test ONLY on FF when we have FF sensors in the 50MP+ range (i.e., FF sensors with roughly the same APS-C crop resolution as the best APS-C sensors).

Direct link | Posted on Dec 8, 2014 at 16:07 UTC as 17th comment | 4 replies
On High-end full frame roundup (2014) article (541 comments in total)
In reply to:

StephenL: When will people realise there is no such thing as full frame? m4/3 is full frame to an Olympus user. Yet to a Pentax 645 owner, a Canon 5D sensor is tiny.

"Full frame" is NOT 35mm. 35mm actually refers to the film gauge popularized by movies (which were then shot 4-sprockets/frame). My understanding is that the use in still cameras began for testing exposure and processing, and then took off as a way of leveraging movie film for stills. A variety of movie film formats, including ones that specify anamorphic compression, are still based on 35mm film.

What still photographers now call full frame is the 36x24mm, 8-sprocket/frame, use of 35mm film in 135 cartridges. So-called half frame 135, at 4-sprockets/frame, was also reasonably common and roughly matches APS-C -- but it is essentially the originally intended 35mm movie frame size. Thus, one could argue 36x24 is actually double-frame 35mm. ;-)

Odd that people get so upset about a name. Also interesting that the most popular still formats today come from a fusion of movie and still technologies that happened a century ago....

Direct link | Posted on Dec 8, 2014 at 15:51 UTC
On High-end full frame roundup (2014) article (541 comments in total)

Let's see... pick a body using a Sony sensor. I don't necessarily agree on the details, but that sounds right to me. Canon really needs to get it's act together.

Direct link | Posted on Dec 8, 2014 at 11:55 UTC as 59th comment | 2 replies
On Up in the air challenge (11 comments in total)

It's done... and the top 3 are all kite photos!

Obviously, KAP is alive and well in the drone age. :-)

Direct link | Posted on Dec 4, 2014 at 02:23 UTC as 3rd comment
On Kalari Fighters in the Up in the air challenge (6 comments in total)

As challenge host, I don't get to vote... but I'm very happy to see a shot like this. It's really impressive -- and even moreso taken from a kite. I suppose it could have been luck, but I think it's much more likely that the photographer deserves credit for recognizing the potential shot and making it happen at the extreme effort of using a kite rig (and at the water's edge, which adds a little to the equipment risk and constrains the kite control).

Direct link | Posted on Dec 1, 2014 at 03:25 UTC as 3rd comment
On Samsung NX mini Review preview (100 comments in total)

The high-ISO raws looks really impressive. I'd guess Samsung is doing significant cooking of their raws, but it's hard to argue with the results. Samsung is doing a great job on the JPEG processing too. Not a camera that appeals to me, but it's worth noting that Samsung is doing a heck of a lot right on everything in cameras right now....

Direct link | Posted on Nov 29, 2014 at 16:00 UTC as 10th comment | 1 reply
On Enthusiast mirrorless camera roundup (2014) article (300 comments in total)
In reply to:

cxsparc: Very weird, where do I find the A6000? Given its performance is better than the Nikon 1 and at least on par if not better than the M1, and also that for the next part the A7 is supposed to be included, this selection is very odd.

Yeah, as many have pointed-out, DPReview seems to have bent over backward to leave out the Sonys that actually dominate this market sector. I get the feeling they wanted the GH4 to win something... and perhaps it should, but what it should win is something more like best self-contained 4K capable (although I'd probably give that to the NX1, a slighting which I guess DPReview compensated for by doing that "fluffy" video "review").

Direct link | Posted on Nov 28, 2014 at 13:31 UTC
On Enthusiast compact camera roundup (2014) article (185 comments in total)

I think all these comparison articles narrow the field too arbitrarily, but do have some useful content within the group they review. Favoring the LX100 seems to have worked in that it's the one listed as "out of stock" at the end of the article. ;-)

Actually, the $120-more-expensive black version LX100 is in stock; it's just the silver version that isn't. Strange price difference....

Direct link | Posted on Nov 26, 2014 at 13:32 UTC as 36th comment
On Up in the air challenge (11 comments in total)
In reply to:

hindesite: I'm very surprised by this - if you read the popular sites, it seems like everybody is using quadcopters and KAP rarely gets a mention, but here the kite photos far outnumber the others. They are also pretty good quality - for something that for me resembles "fishing" for photos :-)

I wonder if kite photography is really more popular, or are the quadcopter pilots spending all their time learning to fly and repairing instead of taking photos? :-)

I suspect we're seeing three things here:

1. KAP has been around for a lot longer than quadcopters.

2. KAP is hard, and the people who do it are very serious about it. The KAP community is filled with people who have great technical skill and have invested lots of time in developing their KAP skills.

3. Kites can lift heavier camera rigs, and go higher, than most cheap quadcopters. The KAP cameras often have better IQ than the cheap, light, ultrawides typically on quadcopters.

For what it's worth, I have an ARDrone 2.0, and the built-in cameras do quite well in the same way the GoPros do well: nice video, somewhat unimpressive stills. I wasn't good enough (ok, maybe I just didn't put enough time in?) to get much going with KAP, but you definitely can get better IQ.

Direct link | Posted on Nov 26, 2014 at 13:12 UTC
On Up in the air challenge (11 comments in total)

I'm very impressed with the kite (KAP) photos. I expected most entries to be from quadcopters, but that obviously isn't the case. Interesting.

Direct link | Posted on Nov 26, 2014 at 03:25 UTC as 6th comment
Total: 512, showing: 1 – 20
« First‹ Previous12345Next ›Last »