mgrum

mgrum

Lives in United Kingdom United Kingdom
Joined on Jan 14, 2009

Comments

Total: 205, showing: 1 – 20
« First‹ Previous12345Next ›Last »
In reply to:

HowaboutRAW: Good for Sony, now reconsider compressed raws in general.

HowaboutRAW:

"In other words you've made my point for me.

You can easily see skips in jpegs-sometimes even big blue bands in the sky."

You've clearly failed to read/understand what I wrote, JPEG compression does NOT work by skipping values, the banding you're seeing in JPEGs is the result of macroblocking artifacts, Sony RAW compression works in a completely different way.

"You can't readd in the subtracted data, unless you have that data stored somewhere. What you've described is making educated guesses and adding those guesses back in. That can work, but not always."

Again you've completely failed to read or understand what I wrote, I'm offering to take an uncompressed 14-bit RAW file (from a Canon/Nikon camera), reduce it to 11 bits, then compare it to the original, demonstrating that nothing of value is lost when converting from a linear 14-bit space to a non-linear 11-bit space. In this case I have the original non-lossy data, I'm not trying to "readd in the subtracted data".

Direct link | Posted on Apr 24, 2015 at 23:15 UTC
In reply to:

HowaboutRAW: Good for Sony, now reconsider compressed raws in general.

HowaboutRAW:

"So I'm guessing Sony is removing some colour data, skipping steps like a jpeg, and that's where the washed out color in recent Sony raws comes from. There I said it."

Please stop guessing, bringing along prejudices from other fields and listen to people who understand the subject matter. Yes it is skipping levels (unlike JPEG which does not skip levels, it converts the image from a spatial to a frequency domain representation and removes low amplitude frequencies) but mostly in the brightest stop which will be compressed anyway when you covert to a gamma corrected colour space.

Finally this is totally different to the case of organic vegetables, in that it's very very easy to to show by experiment that nothing is visibly lost. I have the source code to DCRAW, it wouldn't take long to alter it to add drop 3 bits from a lossless Canon or Nikon RAW and compare the output with the original. I'll do this asap, so check back here in the next couple of days.

Direct link | Posted on Apr 24, 2015 at 15:11 UTC
In reply to:

HowaboutRAW: Good for Sony, now reconsider compressed raws in general.

naththo:

Firstly there is no colour data in a RAW file, there are intensities that are interpreted as colour by the RAW converter, you can't say how many shades you lose because that depends on how the conversion works, but the most naive RAW converter could generate 8.5 trillion colours from 11-bit RAW values, more than the eye can distinguish and more than your display is capable of in any case.

And secondly 3 bits is not a lot to lose as it's mostly from the highlights where variations of 1 level are invisible to the eye. I explained this in detail above, but here it is again: it's not simply chopping three bits off every value, it is transforming values using a curve. In the shadows, where it really counts you're only losing 1 bit (which is likely to be 1 bit of noise in any case). The amount of quantisation increases with brightness, in line with the decreasing ability for people to distinguish brightness levels.

Direct link | Posted on Apr 24, 2015 at 14:28 UTC
In reply to:

HowaboutRAW: Good for Sony, now reconsider compressed raws in general.

naththo:

It's not that bad when you understand a little bit about how images are stored. The eye is far more sensitive to slight brightness variations in the shadows than it is to similar variations in the highlights. Most colour spaces (sRGB, Adobe RGB etc.) take advantage of this fact and store images in a gamma corrected format, this means the brightness is non-linear, it follows a curve with brightness steps starting very fine and growing more coarse as you approach the other end of the scale, mimicking the eyes ability to distinguish similar tones.

Camera sensors are more or less entirely linear, and so is the data coming from the ADC. Sony's compression follows exactly the same process to compress 14 bits to 11 bits using a curve so that more bits are used in the shadows where they're really needed, decreasing gradually toward the highlights.

tl;dr the three bits you lose are not "valuable" they're just storing highlight noise that is invisible to the human eye.

Direct link | Posted on Apr 24, 2015 at 11:35 UTC
In reply to:

HowaboutRAW: Good for Sony, now reconsider compressed raws in general.

HowaboutRAW:

Lossy compression accounts for the "furry" edges to high contrast boundaries, but does absolutely nothing to the colours across the whole image, please stop repeating this falsehood.

Direct link | Posted on Apr 24, 2015 at 08:24 UTC
On Sony SLT-A77 II Review preview (413 comments in total)

I hate to say it again, but really what is the point of publishing a review a whole year after a camera is released? It's common elsewhere to have reviews written before a product hits the shelves. A year later and everyone who is seriously interested would either have bought it based on other reviews or decided to get something else.

It's a good job dpreview don't review films...

Direct link | Posted on Apr 23, 2015 at 14:32 UTC as 39th comment | 4 replies
In reply to:

Nukunukoo: Just trying to do the math. Most anamorphic lenses would turn a 4:3 ratio into the more conventional 16:9-ish ratio. I may be wrong, that means that if I want cinema aspect, my SLR Magic 1.33x won't do. I need a 2x for that.

You could sell a kidney for an Iscorama 1.5x which gets you a little closer... but better to wait for the SLR magic 2x!

Direct link | Posted on Apr 14, 2015 at 11:27 UTC
In reply to:

Mssimo: How bad is anamorphic for still photography?

Not great, sharpness suffers quite a bit especially at wide apertures (where you're likely to notice the anamorphic look) with most available adapters (unless you drop some serious money on a Hawk).

They also vignette when you go wider than about 60mm on full frame. You can almost always get better image quality with a wider lens by cropping the image.

I'm not sure what the new SLR magic anamorphot is like with regards to sharpness or vignetting but it's only a 1.3x

Direct link | Posted on Apr 14, 2015 at 11:25 UTC
On Opinion: Why the Canon XC10 is a big deal article (809 comments in total)
In reply to:

FrankS009: In terms of "convergence" I don't see how this is a more revolutionary product than the Panasonic GH4. And the GH4 shoots RAW still images, and still image can be produced from its 4K footage.

Is it the shape that says convergence to the author?

I would hate to have to the journalist who has to lug 2 lbs around to every event.

Is it the best bang for the buck for companies? We shall see it if sells.

With a 12 megapixel 1" sensor, this camera has the skills capability of a Nikon J1, an entry level mirrorless camera from 2011.

Convergence will happen when you can read a 6k sensor at video framerates (so you can offer competitive resolution in stills mode) with a large enough sensor to give some subject separation (Super35/APS-C), in a body that is ergonomic for both, at a reasonable price.

Direct link | Posted on Apr 9, 2015 at 10:13 UTC
On Sony FE 28mm F2 samples gallery posted article (96 comments in total)
In reply to:

KL Matt: 33 images and not a single one tells us anything much about this lense's bokeh. I don't know about you, but if I'm gonna plunk down nearly 500 bucks for a fast prime lens, I'm absolutely going to want to know how it renders out-of-focus areas before I make a purchase. Yes, even if it's a wide angle, and yes, even if I do need to place my camera close to a foreground object. And yes, I do shoot that way with my wide angle lenses in the field and it's fun and can result in very pleasing and interesting images.

Thematic:

That's nonsense, 28mm is perfect for environmental portraits where you want to maintain some context in the background but have separation from the subject. 50mm lenses wont do this.

Secondly thanks to the full frame sensor you don't have to get that close to throw the background out of focus with a 28mm f/2, certainly not close enough to get any noticeable perspective distortion.

Example:

http://www.mattgrum.com/fm/28bokeh/_DSC6147_small.jpg

Direct link | Posted on Apr 6, 2015 at 10:56 UTC
In reply to:

Atomic Ant: How is it possible to make a lens for all those mounts..but i have to say excellent job.i hope other 3rd party lens makers learn form samyang/rokinon and make their lenses for other mounts also,expecially sigma

Keep the same optics and add a different rear section to the lens barrel that is the correct length and has the correct bayonet fitting for each mount. Simple.

Direct link | Posted on Apr 3, 2015 at 08:27 UTC
In reply to:

tkbslc: Can't figure out why I'd buy this given that nearly every system has a great native 100mm macro for about the same price.

You can't say anything until we know how it performs optically, it's better than the manufacturer's lens at the same price, and offers a better expedience in some areas (dedicated manual focus helicoid, gear teeth for a follow focus) then there is a reason to buy the lens...

Plus not all systems have an affordable ~100mm macro, e.g. Sony FE.

Direct link | Posted on Apr 3, 2015 at 08:22 UTC
On Sony sells half of its Olympus stake article (90 comments in total)
In reply to:

Juck: They got IBIS tech,, the only thing of value in this old girl,, bailed,,, made a profit,, and still have enough shares for a seat on the board. Nice.

@Stu 5

There was a lot of this confusion going around when the A7II was released, essentially:

1) With adapted lenses you only get 3-axis (pitch/roll/yaw) correction. This is because in order to [attempt to] correct for camera X/Y camera translation you need to know the focus distance (objects at different distances move different amounts across the image plane when translating the camera), which is unavailable unless you have a native lens feeding this information back to the camera.

2. With IS enabled native lenses, pitch/yaw correction is done using the lens, whilst roll & camera X/Y translation are corrected using the sensor. Presumably dedicated in lens pitch/yaw correction works better as it's optimised for the focal length(s) of the lens.

3. With a native, non-stabilised lens, you get 5 axis stabilisation from the sensor.

Direct link | Posted on Apr 2, 2015 at 12:03 UTC
On Sony sells half of its Olympus stake article (90 comments in total)
In reply to:

Juck: They got IBIS tech,, the only thing of value in this old girl,, bailed,,, made a profit,, and still have enough shares for a seat on the board. Nice.

@Stu 5

"Sonys system uses a combination of camera and lens to achieve 5 axis"

That's not true at all, you get all 5-"axis" (degrees of freedom, really) by moving the sensor, just like the Olympus system.

Sony claim theirs is developed in house, I see no specific reason to disbelieve that. If anyone else develops a 5-axis sensor stabilisation system it will also work in the same way as the Olympus/Sony systems, because that's the only way to do it.

Direct link | Posted on Apr 2, 2015 at 08:23 UTC
On Nikon 1 J5 offers 20.8MP BSI sensor and revamped look article (152 comments in total)
In reply to:

Sangster: Retro design is nice except proportionally the glass on that kit lens looks a bit small relative to the body.

Can't wait to see the test results on the new BSI 1" sensor.

Yeah I would be surprised if this sensor was from Sony, given how much it has in common with the Aptina sensors in previous Nikon 1s.

Direct link | Posted on Apr 2, 2015 at 08:13 UTC
On Nikon 1 J5 offers 20.8MP BSI sensor and revamped look article (152 comments in total)

It would look much better if the inside of the lens barrel front was black instead of silver, Fuji got it right with the X100.

Direct link | Posted on Apr 2, 2015 at 08:11 UTC as 38th comment
In reply to:

Boss of Sony: Would this lens be good for photographing birds in flight? I'm thinking the wider angle will enable me to photograph birds with a larger wing span?

I got a great photo of a bird in flight with the Canon 10-22mm at 10mm once. Completely by accident, was shooting a seascape when a seagull flew right in front of the camera...

Direct link | Posted on Mar 24, 2015 at 15:32 UTC

"A lower-cost full-frame DSLR with a sensor approaching the same resolution as the 645Z that can accept 645 lenses (via an adapter, and with a crop factor) might well prove appealing to a certain constituency of these users."

There is no "crop factor" when using medium format lenses on a full-frame DSLR with an adapter. Stick a 35mm 645 lens on a DSLR and you will get exactly the same field of view as a native 35mm lens.

Direct link | Posted on Mar 9, 2015 at 11:25 UTC as 30th comment | 1 reply
On Sony Alpha 7S Review preview (478 comments in total)
In reply to:

mpgxsvcd: Wow, Rishi! What a very thorough review. The dynamic range data was very intriguing and not at all what I would have expected. Great job uncovering that.

I still like the A7s because its video capabilities are outstanding. However, I have a lot more respect for the A7r now after this review.

I know you didn't claim to discover it Rishi, that part was addressed to mpgxsvcd who suggested you had. Also I wouldn't say what I wrote was a "slap in the face", it was criticism, you could go as far as to say it was constructive criticism as I explained what I thought was wrong and even suggested an alternative ;)

It was a really good review, it just would have been that much more awesome last year.

Direct link | Posted on Mar 6, 2015 at 14:12 UTC
In reply to:

paul simon king: I have never got over the names Zeiss uses: "Distagon' sounds like some primary school nomenclaure - and just has me making up comic alternatives in my head: "distawent', 'wideolux', 'scopizoomitoff', 'yondergon', splenomegaly - oh no sorry don't know how that last one crept in there
;)

Zeiss terms refer to the optical design, whereas Leica terms just reflect the maximum aperture, which is already part of the lens name: "Oh the 50mm f/2 Summicron is an f/2"!

"The Zeiss 55mm f/1.4 is a Distagon, that's an unusual design for a lens of that focal length, I know something more about it now".

Direct link | Posted on Mar 6, 2015 at 13:40 UTC
Total: 205, showing: 1 – 20
« First‹ Previous12345Next ›Last »