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: 1361, showing: 1 – 20
« First‹ Previous12345Next ›Last »
On article Opinion: DJI has abandoned professionals (287 comments in total)

Two things:

1. DJI has a corporate culture and history of tightly controlling what users can do with their products. Many companies do, and this is problematic for professional users (and researchers like me). I'm talking mostly about things like not providing an API for use of the onboard sensors, etc., in user-written control software... but there also are privacy concerns when a product "calls home."

2. There are signs of an intense battle coming over how UAVs should be allowed to operate. The US government took a shot at establishing a minimal registration program, but now it's clear we need some controls and unclear how they can be established. I suspect DJI put these flight restrictions in largely as a preventative measure against what they would view as more harmful legislation being passed in the future.

Basically, I think this is completely consistent with DJI's history and might help discourage implementation of nastier regulations or at least get DJI products exempted.

Link | Posted on Jul 23, 2017 at 17:09 UTC as 53rd comment | 2 replies
On article Photo of the week: Torment (58 comments in total)
In reply to:

Dheorl: Am I the only one who prefers it cropped to lose the bottom 1/8th or so?

Actually, I think you can make it a stronger (much more abstract) image by just taking the bottom 2/3. The mountains are kind of "predictable." I think I would have tried a horizontal shot getting the "trees" and a little bit of the water pattern as "sky." Come to think of it, having the very bottom a bit out of focus might also have helped... I think I would have tried that too.

Link | Posted on Jul 22, 2017 at 20:29 UTC

"This system has been used only in fresh water...never salt." Hmm. The photo shown above doesn't look like fresh water to me, but maybe brackish...? There are kind-of strange vertical marks on the example shot too. Anyway, I wouldn't try this: water seals must be really dry when the camera is closed, and opening the back to change to a new sheet of film would tend to get the seals moist enough that I'd be worried about leaking if a new sheet was inserted to take another shot before the seals completely dried.... Besides, this is DPReview, right? This isn't D. ;-)

Link | Posted on Jul 22, 2017 at 00:24 UTC as 3rd comment
In reply to:

paul_kew: 8/10, got 4 & 6 wrong. Unfortunately I can't remember which photos they were, I wish they had a more detailed answer page.

I got 7/10, but that was mostly guessing what they wanted me to see. I think they wanted us to see things. However, I honestly feel EVERY photo has been heavily manipulated -- they sure don't look like good quality OOC JPEGs, but softer and artifacted as though they've been JPEG recompressed multiple times, and I found that more distracting than the intended "where's Waldo?" game.

Link | Posted on Jul 21, 2017 at 00:11 UTC
In reply to:

rolleiflex: Similar mistake canon made with lack of c-log on the early batches 5dmk4, which eventually needed a hardware fix. Now canon think they can sugarcoat their turd with a flippy screen and dualpixel AF. It does not work on me.

Im living in a fantasy land where I hope canon are embarrassed enough with headlines on key websites regarding the abysmal DR in this 6D2 sensor, that somehow they can offer a fix to give DR that is at least parity with the 80D. Im not even asking for better, just parity please.

There are at least a couple of tricks possible to improve DR in software. If you can get a dual-pixel raw, there can be about 1 stop extra DR available (not quite, because Canon's raw adds the two sides rather than averaging them -- so it clips). There's also a chance the Magic Lantern folks could do better in camera, not just using dual pixels, but also using the dual ISO hack they created for earlier Canons... but porting ML takes time and that's a development project without the support it so richly deserves.

Link | Posted on Jul 20, 2017 at 12:10 UTC

Does the 6DII have the ADCs on the sensor chip? If not, well, there's your problem.

Canon is still making LOTS of sensors that don't have on-chip ADC, it's just unfortunate if they're reserving on-chip ADC tech only for a few models. In that context, it would actually be performing pretty well for the older tech doing dual pixels....

Link | Posted on Jul 20, 2017 at 05:03 UTC as 327th comment
On article This $31 lens will turn any room into a camera obscura (60 comments in total)
In reply to:

The Sage Knows: In case you're ever in Scotland, you may be interested in:
http://www.camera-obscura.co.uk/camera_obscura/camera_obscura.asp
https://www.visitscotland.com/info/see-do/camera-obscura-and-world-of-illusions-p245621

Well that's cool. Reminds me of http://3dmuseumofwonders.com/

Link | Posted on Jul 19, 2017 at 23:34 UTC
In reply to:

haylebob: dont understand why pic 7 won...can anybody please explain what the photographer means ...

More visceral reaction: Enter play mode, delete image, confirmed.

It has a touch of the "where's Waldo" thing going for it in the range of stuff going on in the image, but what is it? I don't care, so that's a fail. 3 is pretty pointless too -- don't think there even is a Waldo hiding in that one.

Link | Posted on Jul 19, 2017 at 05:30 UTC
In reply to:

OlyPent: You know how it is, sometimes you end up with 1.0, sometimes with 1.1. Depends if the grinding machine is finicky that day or not.

Actually, depends on how you label it. However, I'd bet this is simply a different prototype that maybe they'll make for sale, maybe not.

Link | Posted on Jul 18, 2017 at 04:29 UTC
In reply to:

Sangster: Still isn't as fast the the NASA Zeiss f/0.7 50mm lens Kubrick used for Barry Lyndon.

It's actually not that hard to find an f/0.75 lens. I got one for something like $25 several years ago. The catch is that it's a Kowa 42mm f/0.75 with no aperture mechanism, no focus mechanism, infinity focus is too close to the sensor for even mirrorless bodies, and it would barely cover a cell phone sensor at infinity. It works ok as a specialty macro lens on APS-C ( http://www.instructables.com/id/Using-Ultra-Fast-Lenses-on-DSLR-Cameras/ ), but it's not awesome.

I have a bunch of fast lenses, and for example my Canon FL 55mm f/1.2 works great on a focal reducer to APS-C (where it's f/0.9), but the truth is that most ultrafast lenses suffer a lot of optical defects -- called "character" if you like them. It's rare that an f/1 will do things qualitatively better than an f/2, but it does give you bragging rights. ;-)

Link | Posted on Jul 18, 2017 at 04:20 UTC
On article Tutorial: How to photograph wine on clear plexiglass (45 comments in total)
In reply to:

ProfHankD: This was ok until the gradient, which looks truly terrible. Very few light sources are single-wavelength emitters (it looks unnatural) and dynamic range in a single color is reduced, causing banding and patterned color variation because the integer channel values can't maintain the precise ratio for the color at different brightness (and use of RGB space makes this worse, because RGB is perceptually non-linear). Try polluting the gradient with a little color noise and then focus blur to get a more realistic and smoother texture to the color.

Also, your mask is a little too sharp, so the bottle just looks as pasted-in as it is. First off, you really want to anti-alias the edges when replacing the background. Secondly, a pseudo-3D mask would allow some of the background color to transfer to tint the scene: the edge highlights of the bottle shouldn't be pure white.

Rensol: Did you watch the video? The background of the final image is a single-color gradient generated in editing. Go back and read what I posted. You're actually in agreement with what I said about "Very few light sources are single-wavelength emitters" -- the point being that generated single-color gradients are very artificial, as well as being highly problematic to represent with fixed-precision color tuples.

Link | Posted on Jul 18, 2017 at 00:00 UTC
On article Tutorial: How to photograph wine on clear plexiglass (45 comments in total)
In reply to:

ProfHankD: This was ok until the gradient, which looks truly terrible. Very few light sources are single-wavelength emitters (it looks unnatural) and dynamic range in a single color is reduced, causing banding and patterned color variation because the integer channel values can't maintain the precise ratio for the color at different brightness (and use of RGB space makes this worse, because RGB is perceptually non-linear). Try polluting the gradient with a little color noise and then focus blur to get a more realistic and smoother texture to the color.

Also, your mask is a little too sharp, so the bottle just looks as pasted-in as it is. First off, you really want to anti-alias the edges when replacing the background. Secondly, a pseudo-3D mask would allow some of the background color to transfer to tint the scene: the edge highlights of the bottle shouldn't be pure white.

Rensol: What do you think a photoshop-generated brightness gradient is?

Link | Posted on Jul 17, 2017 at 22:02 UTC
On article Tutorial: How to photograph wine on clear plexiglass (45 comments in total)
In reply to:

ProfHankD: This was ok until the gradient, which looks truly terrible. Very few light sources are single-wavelength emitters (it looks unnatural) and dynamic range in a single color is reduced, causing banding and patterned color variation because the integer channel values can't maintain the precise ratio for the color at different brightness (and use of RGB space makes this worse, because RGB is perceptually non-linear). Try polluting the gradient with a little color noise and then focus blur to get a more realistic and smoother texture to the color.

Also, your mask is a little too sharp, so the bottle just looks as pasted-in as it is. First off, you really want to anti-alias the edges when replacing the background. Secondly, a pseudo-3D mask would allow some of the background color to transfer to tint the scene: the edge highlights of the bottle shouldn't be pure white.

SmilerGrogan: Any perfect gradient is an issue because, even if you use HDR 32-bit float color in editing, images usually eventually get mapped to 8-bit/channel RGB (or even fewer YUV bits in JPEGs). The more extreme the differences in color channel values, the fewer shades get close to the correct ratios -- so making the gradient closer to gray helps. However, the real key is to just add enough noise to break-up the banding -- think of it as replacing highly correlated noise generated by roundoff with uncorrelated noise of about the same magnitude.

Link | Posted on Jul 17, 2017 at 14:22 UTC
On article Tutorial: How to photograph wine on clear plexiglass (45 comments in total)

This was ok until the gradient, which looks truly terrible. Very few light sources are single-wavelength emitters (it looks unnatural) and dynamic range in a single color is reduced, causing banding and patterned color variation because the integer channel values can't maintain the precise ratio for the color at different brightness (and use of RGB space makes this worse, because RGB is perceptually non-linear). Try polluting the gradient with a little color noise and then focus blur to get a more realistic and smoother texture to the color.

Also, your mask is a little too sharp, so the bottle just looks as pasted-in as it is. First off, you really want to anti-alias the edges when replacing the background. Secondly, a pseudo-3D mask would allow some of the background color to transfer to tint the scene: the edge highlights of the bottle shouldn't be pure white.

Link | Posted on Jul 17, 2017 at 12:33 UTC as 7th comment | 9 replies

Excellent! Although I think too many umbrellas....

Link | Posted on Jul 16, 2017 at 18:04 UTC as 78th comment | 2 replies
In reply to:

Ebrahim Saadawi: After reading for hours up on curved sensors, white papers, ten patents or so and latest productions: I can assurely tell you that this is only advantageous for a single focal length per sensor.

If we design an ultra wide angle lens with a correspondingly curved chip that has the cheapest and least number of elements yet gives edge-edge sharpness (long-story-short a really good cheap small ultra wide made possible due to the sensor curve)

and then want to design an 85mm portrait lens for that same chip, we'll be working with constraints of the sensor curve that's ideal for the ultra wide lens and it wouldn't give an advantage to that lens.

Same story with any zoom lens.

It just doesn't work the way they make it seem (we will make sensors that make your lenses cheap and small) Yes you will, but only one lens for one sensor will be made smaller, compared to an identical flat sensor and an identical flat plane lens.

So is it the future? No. Not unless they cam male sillicon chips with photosites and microlenses plus AA and Infrared filters and glass that can bend all as one unit when given a numerical value in-camera (HINT: they can't and won't be able to with any kind of current modern lithography processes, maybe in 50 years or more)

So useless? Of course not: This is, if cheap enough, can make wonders with prime fixed-lens cameras like the Fuji X100 or Sony RX1 or mobile phones or maybe best of all a 16-18mm FF equivalent pocket camera with quite a large chip and edge to edge sharpness, wouldn't that be cool? It would.

Enginel: Interesting. I'd expect that for simple, roughly symmetric, designs around a normal focal length, but not for retrofocus nor telephoto ones. In any case, I don't think the sensor can curve much -- mostly for manufacturing constraints, but also for positioning in the camera. For example, to match curvature to focal length, a non-retrofocus 24mm lens on a 43mm diagonal FF sensor would have the sensor corners lifted nearly 24mm from the center, interfering with lens mounting and making a focal plane shutter impossible.

Link | Posted on Jul 16, 2017 at 13:04 UTC
On article Sphere of frustration: Nikon KeyMission 360 review (202 comments in total)

Well, I got two of these along with a bunch of three other types of 360-degree cameras. This is the most expensive by more than 2X and the worst in every dimension but build quality. Even things like the underwater lens covers are badly designed -- they produce images that look like they were shot through two square portholes. The views are stitched in-camera, but not well, and edge IQ is terrible. Exposure dynamic range seems very limited and it produced video with a black square center when I had the front pointing toward something significantly less bright than a portion of what the back saw; it looks like they lock exposure together and err towards underexposure. Of course, the lack of any status display combined with the ornery wifi compounds all problems.

I was going to post a review, but DPReview doesn't seem to have the entry up for this camera yet, and it also doesn't really fit the Nikon forums, so, here's my very brief review:

RUN AWAY!

Link | Posted on Jul 15, 2017 at 14:56 UTC as 4th comment
On photo Oculus in the Not Rectilinear, But Not Looking Unnatural challenge (10 comments in total)

For what it's worth, the top three are all good, but I personally like #2 best. That tree is haunting... and I love the way everything flows. Also, #1 and #3 take the easy way out in that they don't have things all that close to the camera, while the distortion in #2 really strengthens the composition.

Link | Posted on Jul 14, 2017 at 13:13 UTC as 1st comment

And I thought our camera arrays were awkward to use....

Link | Posted on Jul 13, 2017 at 23:23 UTC as 20th comment
In reply to:

Ebrahim Saadawi: After reading for hours up on curved sensors, white papers, ten patents or so and latest productions: I can assurely tell you that this is only advantageous for a single focal length per sensor.

If we design an ultra wide angle lens with a correspondingly curved chip that has the cheapest and least number of elements yet gives edge-edge sharpness (long-story-short a really good cheap small ultra wide made possible due to the sensor curve)

and then want to design an 85mm portrait lens for that same chip, we'll be working with constraints of the sensor curve that's ideal for the ultra wide lens and it wouldn't give an advantage to that lens.

Same story with any zoom lens.

It just doesn't work the way they make it seem (we will make sensors that make your lenses cheap and small) Yes you will, but only one lens for one sensor will be made smaller, compared to an identical flat sensor and an identical flat plane lens.

So is it the future? No. Not unless they cam male sillicon chips with photosites and microlenses plus AA and Infrared filters and glass that can bend all as one unit when given a numerical value in-camera (HINT: they can't and won't be able to with any kind of current modern lithography processes, maybe in 50 years or more)

So useless? Of course not: This is, if cheap enough, can make wonders with prime fixed-lens cameras like the Fuji X100 or Sony RX1 or mobile phones or maybe best of all a 16-18mm FF equivalent pocket camera with quite a large chip and edge to edge sharpness, wouldn't that be cool? It would.

drajit: the matching is not of focal length, but of curvature of field (except on the sensor side of the lens; curvature of what would be the image plane if it was flat), which is not directly related to focal length in a complex lens.

Link | Posted on Jul 13, 2017 at 13:41 UTC
Total: 1361, showing: 1 – 20
« First‹ Previous12345Next ›Last »