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: 473, showing: 1 – 20
« First‹ Previous12345Next ›Last »
On Canon announces EOS C100 Mark II article (179 comments in total)

This is an unimpressive upgrade of a kludge of a camera. On the bright side, it does sound like perhaps we'll be seeing OLED displays on some Canons now, and better wifi tether support too. The dark side is that the C100 seems to be Canon's idea of what a serious mirrorless body looks like. The C100 was in 2012, then they added a dual pixel upgrade in 2013, and now this. That's significantly more attention than they've given the EOS-M....

Direct link | Posted on Oct 23, 2014 at 02:58 UTC as 49th comment
On Pentax launches K-S1 Sweets Collection article (193 comments in total)

This has got to be a stocking nightmare for stores... but maybe that's the point? Pentax isn't in most stores, and these colors give people a reason to special order while providing a subset small enough for large stores to stock so people can see how pretty they are in person. Not my thing, but hope it works for them... Pentax would be missed.

Direct link | Posted on Oct 22, 2014 at 23:25 UTC as 83rd comment
On Lomography adds Lomochrome Turquoise film to lineup article (71 comments in total)
In reply to:

ManuelVilardeMacedo: I love film but this is not for me. Even Lomography's best efforts to sell a serious film roll are a flop. I tried Lomography Earl Grey, an ASA 100 film, and it has the grain of an ASA 400 film, no sharpness and no contrast. I won't even touch their colour films, especially this one.
However, I do sincerely admire the role Lomography and lomographers have been playing in keeping film alive. It's just that I don't share their viewpoints about how a picture should look like. Shifting colours can be nice, of course, but only occasionally. I don't dig lo-fi.

"I don't dig lo-fi." -- it's worse than that: a fixed unrealistic color rendering is no more artsy than no color shift. In fact, using "good" film, it's pretty easy to implement custom color shifts that are controllable to achieve the specific artistic effect you want (and it's very easy digitally). I look at this more like trying to sell painters on the idea that if they only use two paint colors, a "golden" yellow and "blue" (which sure looks like cyan to me), their painting will be better art. I suppose it looks "novel" the first time you see that color shift, but that doesn't make it art.

Making good art is about being able to implement your artistic vision using the medium you selected. I don't see this film as helping that process in any way; it's providing an inferior medium that offers little artistic control.

Direct link | Posted on Oct 22, 2014 at 00:32 UTC
On Tune in to DPReview Live this Thursday and Friday article (238 comments in total)
In reply to:

Kevin Coppalotti: Dynamic range.want.more. any thing coming?

@Rishi: Shot Noise and desired sample accuracy determine the threshold. Bright pixels may be sampled many times within an exposure interval; dark pixels might not be sampled even once, but we still get a value for them by taking the waveform average over that interval (the smooth waveform might not have any data points within the interval, but is temporally interpolated from samples around that interval). In effect, you get "perfect" data except for some temporal ambiguity, and we're interpolating temporally, perhaps even using super-resolution techniques in time domain.

The target sampling rate is >= 1MHz, to give reasonable S/N for 1/1000s virtual exposures. Fast pulsed lighting is problematic, although many modern strobes are slower than 1/1000s. Flickering lights are also a problem, because flicker makes temporal compression of the waveforms computationally harder or less effective.

This is currently all University of Kentucky research, nowhere near a product....

Direct link | Posted on Oct 21, 2014 at 11:03 UTC
On Tune in to DPReview Live this Thursday and Friday article (238 comments in total)
In reply to:

Kevin Coppalotti: Dynamic range.want.more. any thing coming?

This is getting to be a very long thread.... Anyway, the 13 Ev target has nothing to do with Sony (although they happen to be there at low ISO), it's roughly the upper bound on instantaneous human vision DR. For example, my A7 can "see" more tonal range than I can... and is the first camera I've owned for which that's true.

As for my research, it's a long story, but the relevant thing here is TDCI (time domain continuous imaging) in which each pixel ideally has a processor under it independently timing how long it takes to reach a charge threshold. The sensor output is a (lossless compressed) continuous waveform per pixel describing how light level varies over time. Virtual exposure intervals are then specified after capture, and an image is synthesized by computing the average value of each pixel for the given interval. The idea is that DR is primarily limited by timing precision and range, not by handling of analog charge with all pixels forced to have the same integration interval.

Direct link | Posted on Oct 21, 2014 at 05:10 UTC
On Tune in to DPReview Live this Thursday and Friday article (238 comments in total)
In reply to:

Kevin Coppalotti: Dynamic range.want.more. any thing coming?

@peeve1: No, the DxO measurements don't support the claim that any micro4/3 has "leapfrogged" recent Sony APS-C sensors (let alone FF). They do seem to have edged-out Canon's best.

As for the NEX-7, it's no longer class leading, but it is close despite being quite an old design. It's mostly newer Sony sensors at the top. Samsung looks like the sensor maker most likely to challenge Sony in the immediate future.

In any case, I'd suggest that 13+ Ev is a good minimum goal for DR based on best-case human instantaneous DR -- it's good to be able to get everything the photographer's eyes see (especially if using an OVF). Still more DR is better and, as I said, I'm aiming for 16-20 Ev in the advanced sensor work I'm doing. :-)

Direct link | Posted on Oct 21, 2014 at 00:22 UTC
On Tune in to DPReview Live this Thursday and Friday article (238 comments in total)
In reply to:

Kevin Coppalotti: Dynamic range.want.more. any thing coming?

@Peevee1: You're right! I guess I was biased by seeing old models still for sale (and owned by people I know). The latest crop of micro 4/3 are actually doing slightly better on DR than most Canon DSLRs. The Olympus OM-D E-M10 doesn't get down to 10 Ev until ISO 800, which is darn close to what the APS-C sensor in my old NEX-7 does. Yeah, the NEX-7 is still around a stop better DR at base ISO and Sony's newer models are a touch better still, but that micro 4/3 improved DR by 2 Ev in one sensor generation is quite impressive....

Direct link | Posted on Oct 20, 2014 at 14:09 UTC
In reply to:

goodgeorge: So I voluntarily reduce the the contrast by adding color filter.

In the time of computers is almost any filter (besides polarisers, gray and (strong) graduated filters) better added in computer.

And if you shoot JPG - sell your SLR and buy a superzoom.

@Giklab: no. Instead of capturing multiple ray angles, you'd need to capture multiple polarization angles. There have been array cameras made with an array of polarizers, but it would be a significant hardware change to the Lytro camera. In fact, an ordinary camera with the CFA replaced by a polarizer array is really what we're talking about... although the processing needed vaguely resembles what Lytro does.

Direct link | Posted on Oct 19, 2014 at 11:25 UTC
In reply to:

goodgeorge: So I voluntarily reduce the the contrast by adding color filter.

In the time of computers is almost any filter (besides polarisers, gray and (strong) graduated filters) better added in computer.

And if you shoot JPG - sell your SLR and buy a superzoom.

Drat. Now StevenE's made it a challenge.... ;-)

There are some higher-level vision things that could be tried to recognize reflections, but that's somewhere in OpenCV or MatLab territory, so I'm not really the guy for that....

Direct link | Posted on Oct 18, 2014 at 21:17 UTC
In reply to:

goodgeorge: So I voluntarily reduce the the contrast by adding color filter.

In the time of computers is almost any filter (besides polarisers, gray and (strong) graduated filters) better added in computer.

And if you shoot JPG - sell your SLR and buy a superzoom.

@Wye: Polarization angle is a light attribute that conventional sensors generally cannot distinguish, so, unlike color information, there is no data from which to perform a computational reconstruction. The closest one could get to faking polarization would be tonal mapping based on scene recognition; that could intensify blue skies realistically, but polarizers are often used for removing reflections, and I don't know of any good algorithms for faking that.

BTW, as I've mentioned before, faking film tonal scale and even color sort-of works, but sometimes not to the level of precision that critical cinema users require because RGB CFAs don't provide precise enough spectral data.

Direct link | Posted on Oct 18, 2014 at 15:44 UTC
In reply to:

goodgeorge: So I voluntarily reduce the the contrast by adding color filter.

In the time of computers is almost any filter (besides polarisers, gray and (strong) graduated filters) better added in computer.

And if you shoot JPG - sell your SLR and buy a superzoom.

Polarizers can't be faked in post, but faking a graduated ND is simply a matter of having enough dynamic range. Anything up to a few stops probably can be faked (it would just bring Sony DR down to Canon levels), and using multi-frame HDR capture to fake graduated NDs in post gives far more precise control over the effect than an optical filter would.

The only argument I can see for a tinted filter is to shift the spectral sensitivity either to improve capture of specific bands (e.g., NIR) or to equalize the RGB channels to slightly improve DR. I don't think either is the intent here. I also am not impressed by the cost, although I do appreciate that building a good filter isn't cheap.

Direct link | Posted on Oct 18, 2014 at 11:11 UTC
In reply to:

mpetersson: Is is safe to switch mounts like this? Or is there a point to the current design? Not that I can see any real point to having wiggle-room in the mount, but I'm not an engineer. I have some wiggle with my lenses, but it doesn't really seem to affect anything.

As an engineer, and knowing how careful Sony engineering generally is, I have to wonder if the tiny amount of "give" isn't there to protect against deforming something permanently if a mounted lens is bumped. I haven't been bothered by it on my A7. I suspect the one-piece mount is actually simpler and cheaper than what Sony uses. That said, this replacement is probably not harmful in itself and if it makes people feel better about their camera....

Direct link | Posted on Oct 18, 2014 at 10:49 UTC
On Panasonic Lumix DMC-GH4 Review preview (417 comments in total)

Really a very impressive camera, especially for video -- which seems to be where u4/3 shines. However, the JPEGs don't seem to be as good as the raw would allow them to be, and even raw resolution per pixel isn't awesome... perhaps this is using a heavier anti-alias filter than most cameras? The still image IQ is definitely down a couple of notches from a much cheaper Sony A6000 (which DPReview rated 5% lower), especially for JPEGs, and the price-competitive A7 blows it away.

Direct link | Posted on Oct 17, 2014 at 12:57 UTC as 40th comment | 7 replies
On Canon introduces new $78K 50-1000mm cine lens article (163 comments in total)

Not something I'd be buying, but it's good to see Canon doing something innovative. It does make the 4-pound Sigma 50-500mm f/4-6.3 EX DG HSM look small, light, fast, and cheap, but cinema lenses are never cheap and this does have 2X the focal range of the "Bigma" not counting the teleconverter. ;-)

Direct link | Posted on Oct 17, 2014 at 00:19 UTC as 72nd comment | 4 replies
On Tune in to DPReview Live this Thursday and Friday article (238 comments in total)
In reply to:

Kevin Coppalotti: Dynamic range.want.more. any thing coming?

I understand that light sources blowing out isn't a major concern for most photographers because their own eyes see those things blown, but it's worth noting that it is a huge issue in specific camera uses. For example, automotive cameras that get blinded by sun or headlights are not ok, and a lot of computational "relighting" is based on recovery of environmental luminance maps (capturing the color and brightness of light sources), so both critically depend on high DR.

I think DR is the single most important sensor characteristic. At least from the A100, Sony has given priority to DR; for several years now they've been making sensors with excellent characteristics and DR roughly matching that of human eyesight in good lighting. With the A7S they're close to matching it in poorer lighting. Canon's sensors don't come close unless you use the ML "dual ISO" trick (so it isn't shot noise for them), and quite a few u4/3 sensors are still around 10 Ev max DR.

Direct link | Posted on Oct 16, 2014 at 12:22 UTC
On Tune in to DPReview Live this Thursday and Friday article (238 comments in total)
In reply to:

Kevin Coppalotti: Dynamic range.want.more. any thing coming?

Re: "real scenes having 16-20 EV DR" -- you don't see 16+ Ev DR in all scenes, but even 20 isn't a real upper bound. In fact, if your eyes can see everything simultaneously, it's less than 14 Ev. However, shooting from inside a building looking out to daylight pretty easily gets you 16+. The biggest DR I've seen involves photos taken in caves looking out into daylight, night scenes with powerful lights, or photos including the sun. Those last two are fairly common, especially in the form of specular highlights (reflections off shiny objects).

One of the interesting things about DR is that many systems are actually set-up to deliberately clip about 1% of the pixels, and that usually doesn't bother people. For example, people are generally not bothered by a specular highlight from the sun reflecting off a red car ending up as a fully saturated "white" spot, even though we know it should really still be red... probably because those spots exceed human eyesight DR. ;-)

Direct link | Posted on Oct 16, 2014 at 03:00 UTC

Wow, that's a pretty serious color shift! Is either one right?

Direct link | Posted on Oct 15, 2014 at 22:32 UTC as 19th comment | 1 reply
In reply to:

tinternaut: Absolutely beautiful. However, the engineer in me cringed at the thought of a single point of failure (the camera body) on the photography side.

I'm also nervous about taking only one body, especially in areas where crud is likely to land on the sensor... I at least bring a giottos rocket and sensorklear pen. I'm also wondering where he's charging the batteries while camping... solar charger? Anyway, whatever he's been doing has obviously worked beautifully....

Direct link | Posted on Oct 12, 2014 at 13:44 UTC
On Tiny fps1000 high-speed camera boasts 18,500fps article (137 comments in total)
In reply to:

Sdaniella: Casio Exilim Pro EX-F1 Zoom (2008)
Price: US$1,000 at launch
Liveview: Autogain-LV only (not live exposure preview)

Stills: 2816x2112 (Max hi-cont: 60fps; Max shutter: 1/40,000s)

Video:
H.264/AVC
1920 x 1080 @ 60 fps
1280 x 720 @ 30 fps
512 x 384 @ 300 fp
432 x 192 @ 600 fps
336 x 96 @ 1200 fps
640 x 480 @ 30 fps

Casio Exilim EX-FH25 Zoom (2009)
Stills: 3648 x 2736
Video:
Motion JPEG:
640 x 480 @ (120, 30fps)
448 x 336 @ (30, 120, 240 fps)
224 x 168 @ (420 fps)
224 x 64 @ (1000 fps)

vs
1000fps models:
http://tinyurl.com/1000fps-models

afaik: folks with Canon Powershots using CHDK, can access 1/10,000s shutter speeds and scripts to perfectly time image capture of auto-detecting key high speed short event moments like lightning strikes, etc, stills without ever resorting to hi-speed video.

CHDK is free.
all that remains, like 1000fps, is "time programming" for those who like freeze action capture. such CHDK script only needs to be done ONCE.

Sdaniella, many small sensors in the $50 price range now can do what this project claims using their ROI (region of interest) readout. The difference here is stand-alone camera packaging with dumb dump-to-memory from the sensor's high-speed serial data lines, whereas most competitors do USB3 or GigE tether. It's still interesting as the price is about 1/5 typical high-speed industrial cams and he does say it will be programmable....

The CHDK high-speed video (on a $180 N) looks very programmable, and can probably do much higher framerates than 240FPS, but we don't yet understand the video control registers well enough and the PowerShots don't have high-speed tether nor lots of buffer memory, so we're talking compressed video on them (not raw). BTW, the N is one freaky little camera, hard to recommend for normal users due to a really odd form and UI, but the strange minimalist touch-screen design actually matches industrial-camera type applications quite well.

Direct link | Posted on Oct 10, 2014 at 03:02 UTC
On Tiny fps1000 high-speed camera boasts 18,500fps article (137 comments in total)
In reply to:

Sdaniella: Casio Exilim Pro EX-F1 Zoom (2008)
Price: US$1,000 at launch
Liveview: Autogain-LV only (not live exposure preview)

Stills: 2816x2112 (Max hi-cont: 60fps; Max shutter: 1/40,000s)

Video:
H.264/AVC
1920 x 1080 @ 60 fps
1280 x 720 @ 30 fps
512 x 384 @ 300 fp
432 x 192 @ 600 fps
336 x 96 @ 1200 fps
640 x 480 @ 30 fps

Casio Exilim EX-FH25 Zoom (2009)
Stills: 3648 x 2736
Video:
Motion JPEG:
640 x 480 @ (120, 30fps)
448 x 336 @ (30, 120, 240 fps)
224 x 168 @ (420 fps)
224 x 64 @ (1000 fps)

vs
1000fps models:
http://tinyurl.com/1000fps-models

afaik: folks with Canon Powershots using CHDK, can access 1/10,000s shutter speeds and scripts to perfectly time image capture of auto-detecting key high speed short event moments like lightning strikes, etc, stills without ever resorting to hi-speed video.

CHDK is free.
all that remains, like 1000fps, is "time programming" for those who like freeze action capture. such CHDK script only needs to be done ONCE.

I'm actually doing high-framerate stuff using multiple PowerShot N under CHDK. The N can do 240FPS by itself at 320x240. There are actually quite a few compact cameras with 120+ FPS....

Direct link | Posted on Oct 10, 2014 at 01:18 UTC
Total: 473, showing: 1 – 20
« First‹ Previous12345Next ›Last »