In-camera processing of long-exposure RAW data

Started Jan 21, 2010 | Discussions
RBarbera Regular Member • Posts: 197
Re: In-camera processing of long-exposure RAW data

Have do you see one of the new NEFs from the D7000? At long exposures there are a lot of hot pixels isolated. It seems to me that something has changed in the way they process the RAW files now in-camera.

Yes you get more noise as you open the NEF, but potentially more detail, because it seems that now the camera isn't messing with median average in each channel.

-- hide signature --

Saludos,
Rafa Barberá

Mark Hollister Regular Member • Posts: 358
Re: In-camera processing of long-exposure RAW data

Check the dates; all of the people who contributed to this thread have been dead for hundreds of years now.

I wonder if there was ever any progress on this? It would be a crying shame if there wasn't.

RBarbera Regular Member • Posts: 197
Re: In-camera processing of long-exposure RAW data

Mark Hollister wrote:

Check the dates; all of the people who contributed to this thread have been dead for hundreds of years now.

LOL! only from january 2010

Yes I know that this is an old thread... as the problem with NEF files, that was dated at last from the D70 introduction (see Chrisitian Buil astrophotography site). But the D7000 is the first Nikon camera since the D70 that has hot pixels in long exposure RAWs, so it seems that something is different now, and I'm asking about

-- hide signature --

Saludos,
Rafa Barberá

Mark Hollister Regular Member • Posts: 358
Re: In-camera processing of long-exposure RAW data

RBarbera wrote:

Mark Hollister wrote:

Check the dates; all of the people who contributed to this thread have been dead for hundreds of years now.

LOL! only from january 2010

Yes I know that this is an old thread... as the problem with NEF files, that was dated at last from the D70 introduction (see Chrisitian Buil astrophotography site). But the D7000 is the first Nikon camera since the D70 that has hot pixels in long exposure RAWs, so it seems that something is different now, and I'm asking about

just kidding, actually I'm glad this thread has been kept alive. Marianee really has something here, and if her algorithm were to be adopted, Nikon would look that much better. But as Thom has pointed out in the past, Japanese culture is different from western culture, their pride may not allow them to admit they could have done a better job. But if as you say it seems the D7000 shows evidence that Nikon has abandoned their faulty algorithm, perhaps in the end a better one such as Marianne's will be implemented in future firmware or even PP software. I really wish this thread would continue with input from those with the technical expertise to push this issue to a happy ending.

neveruse New Member • Posts: 1
Re: In-camera processing of long-exposure RAW data

D7000's dark frame still have same "artifact", ie. noise in pairs.
http://bbs.astron.ac.cn/viewthread.php?tid=70632&page=3&fromuid=41203#pid724792
(If you can't read chinese, search "D7000" in the page and check that photo)

OP Marianne Oelund Veteran Member • Posts: 7,779
D7000: A slight improvement has been made to the HPS algorithm

RBarbera wrote:

Have do you see one of the new NEFs from the D7000? At long exposures there are a lot of hot pixels isolated. It seems to me that something has changed in the way they process the RAW files now in-camera.

Yes you get more noise as you open the NEF, but potentially more detail, because it seems that now the camera isn't messing with median average in each channel.

It never was "messing with median average." The hot-pixel suppression (HPS) algorithm isn't a median filter.

That aside, Nikon have made a subtle change to their algorithm in the D7000, with regard to the green channels only. They now treat the two green channels as a single channel, so that each green pixel now has 12 nearest-neighbors (instead of 8) to support its value. This means, on average, there will be a little less reduction in brightness of specular highlights (or stars in astrophotos), but for the most part, the damage is still being done.

Note that for the D7000 (and D3x, if my memory serves me correctly), the HPS process isn't enabled until exposure times reach 1 second. Many other cameras, such as D700/D3, apply HPS when exposures reach only 1/4 second.

Why they refuse to simply give us the option of turning HPS off entirely, is beyond me!

OP Marianne Oelund Veteran Member • Posts: 7,779
Even more improvement than I originally noticed

After examining the test files that I have more closely, it appears Nikon are doing more cross-channel comparisons than just G1-G2. I've also found evidence that the red channel is being supported by adjacent green pixel values, and it would be reasonable to expect the same is true of the blue channel.

I'll need to set up a color-balanced light source in order to continue testing - stay tuned for more.

Mark Hollister Regular Member • Posts: 358
Re: Even more improvement than I originally noticed

Thanks, Marianne for staying on this. I'm very interested in seeing where this goes. It's been a while since I read this thread, but I think I remember reading that a fix for this is beyond the scope of a firmware fix, is that right?

Marianne Oelund wrote:

After examining the test files that I have more closely, it appears Nikon are doing more cross-channel comparisons than just G1-G2. I've also found evidence that the red channel is being supported by adjacent green pixel values, and it would be reasonable to expect the same is true of the blue channel.

I'll need to set up a color-balanced light source in order to continue testing - stay tuned for more.

bob elkind Veteran Member • Posts: 5,815
Re: Even more improvement than I originally noticed

Mark Hollister wrote:

Thanks, Marianne for staying on this. I'm very interested in seeing where this goes. It's been a while since I read this thread, but I think I remember reading that a fix for this is beyond the scope of a firmware fix, is that right?

The fact that HPS is disabled above a certain shutter speed suggests that firmware does indeed control HPS. The enable/disable option which Marianne (and others) would like to use is probably a relatively simple programming addition.

It's less clear that a modified algorithm for implementing HPS would be a firmware-only change. It's plausible that the HPS filter is implemented in the image processing hardware. If so, the chances are greater rather than lesser that a modification of the algorithm would have to be implemented by disabling the filter in hardware, and either filter the images in processor firmware (slower than hardware) or defer HPS to post processing (e.g. CNX2).

This is all wild speculation of course. And Nikon are usually reluctant to share details of sensor structure and shortcomings (e.g. hot pixels) to 3rd parties such as Adobe or Bibble et al.

-- hide signature --

Bob Elkind

Family, mostly sports. Seriously, folks, I'm not that good. If I can do it, you can do it!
photo galleries at http://eteam.zenfolio.com
my relationship with my camera is strictly photonic

 bob elkind's gear list:bob elkind's gear list
Nikon D700 Canon EOS 500D Nikon D600 Nikon AF-Nikkor 80-200mm f/2.8D ED Sigma 12-24mm F4.5-5.6 EX DG Aspherical HSM +5 more
rhlpetrus Forum Pro • Posts: 25,950
D7000 hot pixels

Hot pixels seem more common in the new sensor for D7000. Could it be because of change in this HPS algorithm?

http://forums.dpreview.com/forums/read.asp?forum=1034&message=37043348

bob elkind wrote:

Mark Hollister wrote:

Thanks, Marianne for staying on this. I'm very interested in seeing where this goes. It's been a while since I read this thread, but I think I remember reading that a fix for this is beyond the scope of a firmware fix, is that right?

The fact that HPS is disabled above a certain shutter speed suggests that firmware does indeed control HPS. The enable/disable option which Marianne (and others) would like to use is probably a relatively simple programming addition.

It's less clear that a modified algorithm for implementing HPS would be a firmware-only change. It's plausible that the HPS filter is implemented in the image processing hardware. If so, the chances are greater rather than lesser that a modification of the algorithm would have to be implemented by disabling the filter in hardware, and either filter the images in processor firmware (slower than hardware) or defer HPS to post processing (e.g. CNX2).

This is all wild speculation of course. And Nikon are usually reluctant to share details of sensor structure and shortcomings (e.g. hot pixels) to 3rd parties such as Adobe or Bibble et al.

-- hide signature --

Renato.
http://www.flickr.com/photos/rhlpedrosa/
OnExposure member
http://www.onexposure.net/

Good shooting and good luck
(after Ed Murrow)

 rhlpetrus's gear list:rhlpetrus's gear list
Leica D-Lux (Typ 109) Nikon 1 V1 Nikon D7000 Nikon 1 Nikkor 10mm f/2.8 Nikon AF Nikkor 35mm f/2D +4 more
OP Marianne Oelund Veteran Member • Posts: 7,779
Analysis of new algorithm completed

As explained in my original post on this thread, Nikon's previous HPS algorithm looks at each pixel's 8 nearest neighbors which belong to the same color channel; the pixel being analyzed is reduced in magnitude to the highest value of those neighbors.

If px is a pixel value, and its 8 neighbors have values n1, n2, . . . n8, then the formula for the new px value is
px = Min(px, Max(n1, n2, n3, n4, n5, n6, n7, n8)).

The problem with that algorithm arises since the same-channel neighbors are all 2 photo sites away. Since the immediately-adjacent pixels are not considered (they all belong to other color channels), we end up with such things as small stars with black centers, and faint stars which occupy only a few pixels being completely eliminated.

The new algorithm adds the eight immediately-surrounding pixels to the equation, thus there are now 16 neighbor pixels considered. This will effectively eliminate the black-center problem, prevent loss of faint stars (although their central brightness may still be diminished), and on average, will reduce the impact of the HPS algorithm on the image's small highlights.

Are there any disadvantages to this change? Both versions of HPS are susceptible to hot-pixel pairs or clusters, in that they will fail to remove a hot pixel which has another hot pixel near by. The new algorithm, since it looks at twice as many neighbors, can be expected to leave about twice as many of these hot-pixel pairs untreated. The impact of this will depend on the sensor's propensity for generating hot pixels in closely-spaced pairs.

Besides the algorithm change, there could also be other changes, such as in the way long-exposure NR operates, or the cases for which HPS is enabled. I haven't looked at these possibilities yet, but will post again if I find anything significant.

rhlpetrus Forum Pro • Posts: 25,950
A test

Someone is also looking at that, it seems that Nikon is doing a very good job for astro photog with D7000:

http://forums.dpreview.com/forums/read.asp?forum=1034&message=37060238
--
Renato.
http://www.flickr.com/photos/rhlpedrosa/
OnExposure member
http://www.onexposure.net/

Good shooting and good luck
(after Ed Murrow)

 rhlpetrus's gear list:rhlpetrus's gear list
Leica D-Lux (Typ 109) Nikon 1 V1 Nikon D7000 Nikon 1 Nikkor 10mm f/2.8 Nikon AF Nikkor 35mm f/2D +4 more
bob elkind Veteran Member • Posts: 5,815
Silly post: D7000 hot pixels

rhlpetrus wrote:

Hot pixels seem more common in the new sensor for D7000. Could it be because of change in this HPS algorithm?

http://forums.dpreview.com/forums/read.asp?forum=1034&message=37043348

Renato, this is a silly post. Have you read the thread you linked?

Is this wild speculation on your part, or do you have something serious to say based on facts rather than thin air?

-- hide signature --

Bob Elkind

Family, mostly sports. Seriously, folks, I'm not that good. If I can do it, you can do it!
photo galleries at http://eteam.zenfolio.com
my relationship with my camera is strictly photonic

 bob elkind's gear list:bob elkind's gear list
Nikon D700 Canon EOS 500D Nikon D600 Nikon AF-Nikkor 80-200mm f/2.8D ED Sigma 12-24mm F4.5-5.6 EX DG Aspherical HSM +5 more
rhlpetrus Forum Pro • Posts: 25,950
Re: Silly post: D7000 hot pixels

Sorry, indeed, the person even withdrew the images. But there's another one with a 20 min capped image that showed (see link in post below) pretty low hot pixel effects. That, if confirmed, together with Marianne's findings, might indicate that the D7000 is appropriate for astrophotography, more so than previous Nikons.

Cheers!

bob elkind wrote:

rhlpetrus wrote:

Hot pixels seem more common in the new sensor for D7000. Could it be because of change in this HPS algorithm?

http://forums.dpreview.com/forums/read.asp?forum=1034&message=37043348

Renato, this is a silly post. Have you read the thread you linked?

Is this wild speculation on your part, or do you have something serious to say based on facts rather than thin air?

-- hide signature --

Renato.
http://www.flickr.com/photos/rhlpedrosa/
OnExposure member
http://www.onexposure.net/

Good shooting and good luck
(after Ed Murrow)

 rhlpetrus's gear list:rhlpetrus's gear list
Leica D-Lux (Typ 109) Nikon 1 V1 Nikon D7000 Nikon 1 Nikkor 10mm f/2.8 Nikon AF Nikkor 35mm f/2D +4 more
OP Marianne Oelund Veteran Member • Posts: 7,779
Comparison example: D7000 versus old HPS algorithm
1

This example compares the new D7000 HPS algorithm to the old algorithm, using a test image with a couple of features added. You will see that although there is some loss of brilliance of fine features overall, the D7000 version algorithm doesn't produce nearly as dismal of a result as the old algorithm did.

Test features added to the original image include a set of 3 simulated hot pixels, which are removed by both algorithm versions, as desired. However, a small 2x2 pixel simulated star image is also removed by the old algorithm, showing why it is so disliked by the astrophotography community. The D7000 algorithm successfully leaves this small object intact.

Martin Grecner Contributing Member • Posts: 520
Re: Comparison example: D7000 versus old HPS algorithm

are removed by both algorithm versions, as desired. However, a small 2x2 pixel > simulated star image is also removed by the old algorithm, showing why it is so disliked > by the astrophotography community. The D7000 algorithm successfully leaves this small > object intact.

Wow, I am amazed ! Do I understand it right that you have managed to shoot the test picture so that it is pixel-on-pixel mapped onto the camera sensor ? Is it some sort of trial and error or something more sophisticated ?

Anyway, if you findings are true, then it really can only mean that:

1. D7000 sensor has hot-pixels which appear in clusters (two or mor adjacent hot-pixels), and that is the reason why they can be seen in the final image, since the HPS can not handle such pixels

2. The reason for hot-pixels in D7000 video mode is still questionable, since we do not know how the sensor's 16 megapixels are mapped onto the FullHD video 2 megapixels.

If you can do the same sort of pixel-on-pixel shooting in video mode, it would be more than interesting to see not only how the D7000 handles hot-pixels in video mode, but also how does it downsample the sensor to achieve the FullHD video resolution of 1920x1080 pixels.

Martin

rhlpetrus Forum Pro • Posts: 25,950
Thanks n/t
-- hide signature --

Renato.
http://www.flickr.com/photos/rhlpedrosa/
OnExposure member
http://www.onexposure.net/

Good shooting and good luck
(after Ed Murrow)

 rhlpetrus's gear list:rhlpetrus's gear list
Leica D-Lux (Typ 109) Nikon 1 V1 Nikon D7000 Nikon 1 Nikkor 10mm f/2.8 Nikon AF Nikkor 35mm f/2D +4 more
Dale Ireland New Member • Posts: 14
Re: Comparison example: D7000 versus old HPS algorithm

I wonder why Nikon does not advertise this improvement since it can make the camera so much more popular .

OP Marianne Oelund Veteran Member • Posts: 7,779
Nikon silence

Dale Ireland wrote:

I wonder why Nikon does not advertise this improvement since it can make the camera so much more popular .

Fine points such as this are typically withheld from public release. In this case, it could sound like an admission that the older HPS algorithm wasn't optimal; moreover, they've never even admitted that the HPS algorithm exists! . Often, it's up to independent testers to discover and publicize detail improvements that they make.

For example, consider the vague remarks that Nikon typically make about improvements to the AF system. They will tell us that "AF is improved" by a firmware upgrade, but never go into detail unless there are changes to the user manual. I've found significant functional changes to AF which were never hinted at in their firmware announcements.

Vahur Krouverk Contributing Member • Posts: 659
Re: Comparison example: D7000 versus old HPS algorithm

I wonder whether this changed HPS algorithm is cause for Nikon's firmware fix announcement and with next D7000 firmware "old" algorithm will be used again?

Keyboard shortcuts:
FForum MMy threads