Pixel size and image resolution

[Beatsy]

Patents used to be a good thing. Now they've been spoiled by "squatters". Also known as lawyers, or legal companies, who instigate vexatious litigation against companies who actually figure out how to make stuff work. And all they (the vexatious squatters) ever did was buy the company who's patents were too-easily granted and never should have been, on the grounds of "prior art". The patent system is no longer fit for purpose IMO, it's been hijacked and is constantly abused by those with money and/or bigger lawyers than me.

Did you know that winning a vexatious patent litigation still costs a million quid, even if you win outright!!!! Horse puckey!

[mjkzz]

Yes, you seem to be more concerned with reducing noise, while I am most interested in sub-pixel resolution and complete color sampling. The Oly algorithm does both those things with the minimum possible number of shots. It is a beautiful algorithm.

OK, you brought up the issue of noise reduction and as I said, SR algorithm does that even without pixel shifting, AND SR algorithm does color sampling and all that stuff which you wanted to separate. Oly algorithm is beautiful does not mean the underlying algorithm is NOT a variant of SR.

My point is, I GUESS these manufacturers are hyping a technology that could be well known and well researched, but their sales pitch makes it as some black magic, and in fact, they have achieved it.

Anyways, I think we have a break down of communication, I understand what you are saying and you are right.

[mjkzz]

the key is all about subpixel alignment of images and some trivial manipulation

This is sounding more than vaguely similar to https://petapixel.com/2015/02/21/a-prac ... photoshop/ .

We discussed this a couple of years ago, at http://www.photomacrography.net/forum/v ... 967#183967 and in the surrounding thread, and in the followup thread at http://www.photomacrography.net/forum/v ... hp?t=29822 .

--Rik

After reading that article, I think the problem lies in discretization during the 200% up sampling. For shaky hand method, here are my thoughts:

It is almost impossible to have an integer # of pixel shift because the border of pixel is so thin. So most of images will have fractional pixel alignment.

Shifts in pixels includes translational and rotational, we can probably ignore the pitch and yaw rotation for argument sake, but we can not ignore the roll rotation. It could look something like this:



The problem with resizing up to 200% and then align the up sampled image is that information are not fused correctly due to the manual procedure which is discrete in nature. Therefore, it is reasonable to say "However, the technique of superresolution by random shifting is a different matter entirely -- not nearly as predictable or reliable, and remarkably easy to misinterpret." as evidenced by your illustration of pixelated part.

However, if images obtained by shaky hand are fed into an algorithm where floating point calculation is used, things would be different and I believe by taking more samples, it can produce better SR image because information fusion will be more precise.

[Macrero]

Macero,

Agree, that in the right hands (like yours) and studio work the sensor could really show it's stuff

What do you know about the latest sensor vs. the previous ones regarding performance?

Best,

Mike,

I recently started looking into the Quattro cameras. The sensor of the SD Quattro (not H) is not that new, but it is obviously improved compared with the old Foveon in the SD 14 (2006). Though it seems that there are still remaining some issues that Sigma has not been able to sort out, such as the horrendous high-ISO performance (not a problem in studio), "odd" artifacts/noise (even at base ISO) especially in the out-of-focus areas and in the shadows (this may be/will be a problem).

I'll not know how the camera will perform for stacking till I try it... It will not be a piece of cake, but I like challenges Curiously it seems that nobody has tried Sigma SD cameras for stacking, or at least I haven't found it, on these boards and anywhere...

Best,

- Macrero

[Lou Jost]

It will be very interesting to see!

[RobertOToole]

.......Curiously it seems that nobody has tried Sigma SD cameras for stacking, or at least I haven't found it, on these boards and anywhere...

About 2 years ago Sigma did give me a few of the dp Quattros, then new, for a project, not the current SD Quattros, and I shot with them for the project. As part of the deal I ended up with two of the cameras and all the accessories and tried a stack or two. I sold them soon after. I don't remember anything significant about the image quality other than the cameras were glacially slow. That is the main reason I sold them honestly. I may still have the AML-72 and hood Macrero if you are interested, just pay for postage

BR,

Robert

[RobertOToole]

.....Though it seems that there are still remaining some issues that Sigma has not been able to sort out, such as the horrendous high-ISO performance (not a problem in studio), "odd" artifacts/noise (even at base ISO) especially in the out-of-focus areas and in the shadows (this may be/will be a problem).
- Macrero

In regards to noise, as far as I can remember the Sigma sensor shadow noise improvement was very much dependent on ISO setting so the cameras exposure had to be set for extreme over exposure. Not just expose to the right technique, but extreme expose to the right, not ETTR but EETTR!

The current Nikons and Sony sensors, other than the D5 and A9 are very much ISO setting invariant which is a huge advantage working in the field. (that is setting the ISO higher in camera does not improve shadow noise)

Robert

[Macrero]

About 2 years ago Sigma did give me a few of the dp Quattros, then new, for a project, not the current SD Quattros, and I shot with them for the project. As part of the deal I ended up with two of the cameras and all the accessories and tried a stack or two. I sold them soon after. I don't remember anything significant about the image quality other than the cameras were glacially slow. That is the main reason I sold them honestly. I may still have the AML-72 and hood Macrero if you are interested, just pay for postage

BR,

Robert

Yep, I'm aware of how slow Sigma cameras are that will make stacking a tad tedious, especially given that I work with a manual rail...

I have a lot of stuff, more than I need and use... Thanks for the offer anyway I actually didn't need a new camera either, but I am just curious to see how this "new" Foveon works for studio stacking. I'm by no means sure it will perform flawless, but the only way to find out is trying it. If it don't does well I will return it or sell it later on.

In regards to noise, as far as I can remember the Sigma sensor shadow noise improvement was very much dependent on ISO setting so the cameras exposure had to be set for extreme over exposure. Not just expose to the right technique, but extreme expose to the right, not ETTR but EETTR!

The current Nikons and Sony sensors, other than the D5 and A9 are very much ISO setting invariant which is a huge advantage working in the field. (that is setting the ISO higher in camera does not improve shadow noise)

Robert

I was just playing with the latest version of the Sigma Photo Pro software and some sample RAWs and it does a very good job suppressing the native Foveon's noise and artifacts while preserving detail. Far from what I was getting back in the day from the SD 14 and the SPP version at the time. That's encouraging.

I will try that "not ETTR but EETTR" trick

Best,

- Macrero