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: 682, showing: 41 – 60
« First‹ Previous12345Next ›Last »
On Sources of noise part two: Electronic Noise article (234 comments in total)
In reply to:

junk1: This article matches my recollection from my MSEE classes. I remember doing the math, but can't recall it from 20 years ago.

Noise in the sensor / front end is very bad since if you amplify the signal 100x in the next stage (or "downstream"), that incoming (into the amp) noise is also (of course!) amplified 100x. This is why you want a strong pre-amp signal from your audio system - so it does not need to be amplified as much later "downstream".

The reason CHDK and ML work is that it isn't just the sensor tech that has stayed remarkably constant in Canons for over a decade, but pretty much the entire camera architecture. It's not just the sensor that needs to change in major ways, which makes change a very slow and expensive process. For example, the byte order of raw image data matches old 16-bit hardware, and awkwardly doesn't match the order used by the ARM32 processors in Canon cameras, but this mismatch has persisted for more than a decade (presumably because some of that ancient hardware is still in the JPEG processing pipe).

One has to remember that the camera industry Canon grew up in is one in which product life cycles were measured in decades, with almost all improvements being incremental changes to individual subsystems. They've not yet adjusted to the fact that cameras now are primarily computing devices that need to be designed as tightly integrated systems in short development cycles.

Direct link | Posted on May 15, 2015 at 10:36 UTC
On False Colors challenge (3 comments in total)

A reminder: the rules say you should describe the color change(s) you made. I will not disqualify entries failing to describe the color changes, but strongly encourage voting to penalize them.

Direct link | Posted on May 15, 2015 at 09:36 UTC as 3rd comment

Double reflections were in 96/197 test images -- that's less than 49% that this algorithm has any hope of working on. Still cool that it almost sort-of works with such a computationally cheap algorithm, but would any press coverage have been given this if it wasn't associated with MIT? (Were any previous algorithms?) The article also credits someone at the Hebrew University of Jerusalem as co-developer of the algorithm, yet the article title here says the work comes from MIT. Well, this "news" did come from the "MIT News Office." ;-)

Direct link | Posted on May 14, 2015 at 11:03 UTC as 30th comment | 1 reply
On Sources of noise part two: Electronic Noise article (234 comments in total)

Nice overview. Two notes:

1. JPEG tone curves are not what they used to be. It used to be that you'd never see more than about 9EV dynamic range mapped into a JPEG, but now there are various "fancy" tone-mapping algorithms (e.g., shadow boosting like Sony's DRO) and S-log-type encodings. Put another way, anything you can do to a raw in postprocessing could theoretically be done in-camera, and an increasing variety is being done in camera, especially using CHDK (Canon Hack Development Kit) or ML (Magic Lantern), so camera JPEGs can reveal much more than is suggested here.

2. Electronic noise characteristics vary wildly across cameras, especially when you include webcams -- which are still a large fraction of all cameras out there, although not one of great interest to most DPReview readers. A typical $3 webcam has a very different electronic noise profile from your Nikon D810, not just more noise (e.g., webcams often suffer various types of strongly patterned noise).

Direct link | Posted on May 13, 2015 at 13:57 UTC as 65th comment
In reply to:

terp02andrew: I am definitely intrigued by the 8mm. The Panny f3.5 fisheye I have now isn't bad, but f1.8 is a significant improvement :p The larger aperture does result in the Oly being about twice the weight though.

Looking at the gallery images, the Oly fisheye does look really good: no flare issues (amazing for a 17-element lens!), sharp even wide open, very little CA, only a bit of left-side softness in some images.

Still, I'm somehow not thrilled with the feel of the sample images. I don't know if it's that I'm not used to the 4x3 aspect for fisheyes or if it's a slightly odd projection choice, but something seems a bit off...? Maybe I've gotten used to Samyang's unusual projection?

Direct link | Posted on May 12, 2015 at 13:29 UTC
In reply to:

terp02andrew: I am definitely intrigued by the 8mm. The Panny f3.5 fisheye I have now isn't bad, but f1.8 is a significant improvement :p The larger aperture does result in the Oly being about twice the weight though.

I think f/1.8 is the answer to "why not just get the Samyang?"
BTW, it isn't a particularly great answer. ;-)

Direct link | Posted on May 12, 2015 at 11:12 UTC
In reply to:

davetj_uk: You can get an App called Camscanner for free for android/ipad that will scan in pdf or jpg.. it does a brilliant job as well.. Camscanner straightens your pictures, and crops out any unwanted space.. I wouldn't lug that QromaScan thing around with me to copy photos and I cant believe anyone would want to pay for such a thing.. I could knock that up with a bit of strong cardboard and use tissue paper as a diffuser for a table lamp or natural light.. $35... You got to be kidding me!!

Holding the camera parallel and the LED lighting are the main benefits, but I wonder about the LEDs. It can be very difficult to get even lighting with point sources, and the inside of the box seems to have made bad color choices.

Direct link | Posted on May 6, 2015 at 11:30 UTC
On Big Zoom: Nikon Coolpix P900 real-world samples article (154 comments in total)
In reply to:

brn: Warning: not for pixel peepers.

At 1:1, even this ISO100 images look terrible.

True enough. However, they look consistently good scaled to 4MP (well, with somewhat nervous bokeh) and that's sort-of enough for most purposes. This is about getting the shot with no lens changes, not ultimate IQ.

Direct link | Posted on May 5, 2015 at 23:58 UTC
On A photo. challenge (10 comments in total)
In reply to:

Chris Page: Is this a late April fools?
Seriously...

Glad we can use PS to get rid of the red eye of the Fiji Fish caught on a Friday and taken with a Fish eye lens!

Well, it better be a joke... it just seems annoying. The rule:

"Size on shortest side must be no shorter than 196.4 megapixels."

ensures there cannot be any valid entries, because an image with that dimension would not be accepted by the DPReview image uploader.

I've been very frustrated by how little attention folks pay to the (very reasonable) rules in the challenges I've posted... perhaps this challenge host has been driven insane by such behaviors? ;-)

Direct link | Posted on May 5, 2015 at 11:22 UTC

Ok, this is silly, or sad, or innovatively artsy, depending on how you think about it. However, the photos at the WWW site show the cell phone in the vertical orientation with every photo in the horizontal orientation. Beyond that, you'd think there would at least be something like a remote trigger built-into the hand, but no such thing is mentioned.

In sum, the idea actually has potential as a joke, art, etc., but this isn't a well-thought-out thing.... I predict we'll see a cheap, more functional, copy from China on eBay within a few months. ;-)

Direct link | Posted on May 3, 2015 at 13:08 UTC as 81st comment | 1 reply
In reply to:

SteB: As I've tried to explain on the forum, this is very uniform distribution, which is the opposite of random distribution.
http://en.wikipedia.org/wiki/Randomness

If this was just random dust it would not be distributed like this. It could be dust trapped in the sandwich covering the sensor, but there would have to be something in the manufacturing process causing it to be evenly distributed. This is why I'm guessing a fault in the manufacturing process, for it to cause this uniform distribution.

Anyone who has studied distribution, randomness, statistics and probability in depth, will be aware that it is very unlikely that such a uniform distribution like this would be caused by a random fault like dust falling onto something. It could be dust again, but it would be dust formed by part of a process.

I'm fairly certain this will be fixed, once they discover what part of the manufacturing process is causing it.

As I said below, my guess is a defective sputtering process for applying a coating. What do I win if I'm right? ;-)

Direct link | Posted on May 1, 2015 at 11:03 UTC

Interesting defect; I've not seen anything quite like it. Looks sort of like a very even distribution of tiny water spots (e.g., like fine mist settling on a lens surface)... maybe some kind of sputtering artifact when applying a coating? It will be interesting to see what this really is....

Direct link | Posted on May 1, 2015 at 03:10 UTC as 124th comment | 1 reply
In reply to:

iAPX: Good upgrade. But I think that people interested in it won't like the video or live-view features, and would have preferred a simplified camera, to take pictures the old way.

Monochrome videos at 1080. I think a 4K color video converted to monochrome will beat it easily... assuming anyone cares about monochrome videos.

Direct link | Posted on May 1, 2015 at 00:40 UTC
In reply to:

Mark Banas: Waiting for someone to do the math on the prize value vs. # of prior paid entries. ;)

That said, it's great to have any excuse to shoot more panos, and I see Peter Lik was one of the past judges, while Aaron Spence is one of the current ones. The standards have gone up!

There are several reasons to run a photo contest. There are tons of photo contests that are basically scams in which getting the entry fees is the real goal... I don't think that's what Epson is doing here. For Epson, I'd expect it to be mostly about goodwill and bolstering the brand, but this is Epson Australia, not all Epson, and it seems likely to me that they literally didn't have the cash to invest in this. Actually, it isn't even clear that Epson initiated this; there are other sponsors, so Epson might just get top billing because they contributed the most.

So, a little math says the expected value of entering is approximately loss of 40% of your "investment" in entry fees, and the prizes are mostly not cash (yet probably taxable for the winners). Not compelling... unless you're sure you can win. These are at least way better payoff odds than you'll get gambling in Las Vegas. ;-)

Direct link | Posted on Apr 30, 2015 at 02:54 UTC
On GoPro announces Kolor acquisition article (46 comments in total)
In reply to:

Paul Guba: I am uninspired by the footage.

Yeah, it does lack awesomeness. However, they do a darn good job of hiding where the seams are and how the cameras are being held (they are being held by something, so the removal of that is being done quite well). I'm a little disappointed if it's really 3 GoPros and not one new 360-degree model... but perhaps that's in the works?

Direct link | Posted on Apr 29, 2015 at 18:08 UTC
In reply to:

Photato: About Sensor Size.
Strictly speaking Sensor Size should not be a factor in Noise, because lenses can concentrate the same light (photons) in a large or small area.
So much so that you can start a fire concentrating Photons with a magnifying glass in a small area.

What happens is that normally smaller sensors have a higher Density of pixels making it less efficient in photon collection.

For instance. A Small Sensor filled with 8 Micron Pixels should be able to collect the same amount of Photons than a Bigger Sensor filled with 8 Micron pixels. The difference is that the Bigger Sensor would have higher resolution.

As I was hinting above, quantum efficiency usually isn't the same for large and small sensors. A larger fraction of the photons hitting the smaller sensor get counted, which reduces noise. It is also often true that the maximum charge that can be stored is increased by the more aggressive design rules/ fab technology used for smaller chips. In sum, the advantage for big sensors isn't quite as big as might be expected.

Direct link | Posted on Apr 28, 2015 at 03:32 UTC
In reply to:

Photato: About Sensor Size.
Strictly speaking Sensor Size should not be a factor in Noise, because lenses can concentrate the same light (photons) in a large or small area.
So much so that you can start a fire concentrating Photons with a magnifying glass in a small area.

What happens is that normally smaller sensors have a higher Density of pixels making it less efficient in photon collection.

For instance. A Small Sensor filled with 8 Micron Pixels should be able to collect the same amount of Photons than a Bigger Sensor filled with 8 Micron pixels. The difference is that the Bigger Sensor would have higher resolution.

Actually, smaller sensors tend to be better than larger ones (per unit sensel area). Chip yield drops dramatically with area increase unless you use a more conservative design/fab technology. This partially compensates for the usual benefits of larger sensels.

Direct link | Posted on Apr 27, 2015 at 19:59 UTC
In reply to:

ProfHankD: These look clearly designed for Sony FE bodies, and even the OLED display makes sense for that (the info was digitally available, so I bet this doesn't cost any more than the usual focus/DoF scales). A big step for Zeiss... a giant endorsement for Sony.

Sidath: Yup, and also the fact that they've obviously done some new electronics -- not just basic AF, but also the cute little OLED display on the lens. No, it isn't earth-shaking, but Zeiss is a very conservative company, so I do see this as a big step for them.

Direct link | Posted on Apr 26, 2015 at 16:22 UTC
In reply to:

Eugeniu Sofroni: Canon had this option with the Magic Lantern firmware add-on www.magiclantern.fm/features.html

it also has motion detection trigger, lightning trigger ... many other options

No CANON does not have any such thing -- Magic Lantern has a variety of triggering options, and CHDK has even more, and they happen to work on Canon cameras. Canon's lack of security in their camera firmware allows benevolent open source development, but Canon is in no way supportive of either ML nor CHDK.

In contrast, Sony, which ironically uses open source Linux in all their cameras, uses Linux security features to effectively prevent anyone from developing such goodies for their cameras. Sony is slowly moving toward enabling powerful and portable remote control (via wifi and JSON protocol), but nothing allowing 3rd-part code in the camera yet. I wish they would let us open source developers develop for their cameras, and use the security features to void the warranty if you run a non-Sony-approved app (with an Apple-like approval process for camera apps), but nothing yet along those lines....

Direct link | Posted on Apr 25, 2015 at 17:32 UTC
In reply to:

Marty4650: Those are the best quality photos I have ever seen coming from a smartphone.

Of course, the real question is.... how many smartphone users want, need, or are willing to pay for better quality photos? Over one billion smartphones were sold last year, and very few of them were purchased to take high quality photos.

You really don't need image quality this good for facebook posts or email attachments. So will these users pay a very high premium price to get better photos?

My guess is.... no.

Nice exercise in proving it can be done, but I think Apple and Samsung are safe.

Better IQ in phones is gonna happen. Up to around ISO400, this is darn pretty, if slightly artificial looking. Higher still works. Overall, I think you'd be able to get really nice images scaling down to 5MP, which is plenty for nearly everything. All good.

Will people buy it in droves? Probably not; it isn't from Apple nor Samsung. Still, it is another step in an inevitable direction.

Direct link | Posted on Apr 25, 2015 at 03:17 UTC
Total: 682, showing: 41 – 60
« First‹ Previous12345Next ›Last »