This content has been marked as final. Show 31 replies
Picasa could be reading just the embedded JPEG in your raw image.
Can you post a raw file to an FTP site or to yousendit.com?
I agree - Picasa may be doing that. But, I can also adjust the image too, seemingly without the banding that you'd expect from editing an 8-bit file.
Further, the raw file looks much better in Nikon Capture NX:
The raw file is here:
The photo is underexposed - one of the nice things about RAW is that you can correct that much more than if you were shooting JPG. In any case, the pattern is too regular to be noise from a low-light situation, right? This noise is simply not random.
Compressed NEF of not?
Great question - the NEF above is 2000x3000, 12 bits per channel, and is 5.23MB - doing the math, it seems it must be compressed, though I admit I don't know how to set or unset this on my D70s.
Sorry, NEFs are always compressed on D70, D70S and D50.
What are your sharpening settings in ACR? I don't understand how iPhoto has the same issue.
How are you downloading the images?
Even you Nikon Capture NX processed image looks abnormal. Either you are having a file corruption problem, or your camera is defective.
In all cases, I am using a USB card reader to transfer images. For iPhoto and Lightroom - those applications do the import. For Capture MX, I worked with an already-transferred file. I have also tried plugging the camera in via USB and doing the transfer via the camera, but the effects are the same.
This looks like a defective sensor/camera to me. The NX image is also defective and has a weird grid over it that probably trips up the demosaic in ACR and iPhoto. Picasa often simply reads the embedded jpeg BTW and is useless for any photography work as it does not color manage.
Wow, this is strange. It does not look like normal image noise to me. At 100% there are light horizontal bars and dark vertical bars all over.
I ran it through NeatImage and it cleaned it up a lot. But there is still some visible banding and loss of detail in some skin tones such as the forehead.
It almost looks like something went wrong while the image was being recorded. I encountered something similar with a Kodak body and long exposure in low light. That is not the case here.
Was this shot through some kind of scrim or something in front of your lights?
Cheers, Rags :-)
I have seen this pattern before. A few years ago someon posted a file from an Olympus camera that did the same thing. But I don't remember what the outcome was. I distinctly remember that pattern in there that looks sort of like a maze. But with that one it was only certain raw processors that gave this result.
I recall seeing the same pattern when I first started shooting digital and was making long exposures at low light.I believe it is overexposure. I shoot with a Canon 20D.
Helen - thanks for the feedback, but I don't think that is the case. The exposure on the images linked to above is about 1/1600 sec, and the histogram shows them to be several stops underexposed
Really what you see here is not noise, it is a regular pattern that is overlaid on the image. Such regular patterns, get interpreted in a weird fashion by most RAW converters. You can clearly see it in the Picasa image too. it appears to me a broken amplifier in the camera that is giving interference during the readout. You probably need to send it in for repair.
The "maze" is due to the unequal reponse of the green pixels in the two alternating raws. The following capture shows the green pixels between 290 and 351; the green at the upper limit appear row-wise.
As the image is hugely underexposed and it has to be corrected, this difference becomes somewhat magnified, and the de-mosaicing does the rest.
Using a de-mosaicing algorythm, which differentiates between the two "kind" of greens in the CFA eliminates the problem, but it costs contrast (DCRaw can do that). Better exposure is more of a help.
For the demonstration of the effect of different de-mosaicing algorythms see http://www.pages.drexel.edu/~par24/rawhistogram/40D_Demosaicing/40D_DemosaicingArtifacts. html
G Sch - thanks for the response. It gives me hope that the camera and my typical PP setup is fine. Now, I guess I need to hit the books to fully understand the technical details of what you are talking about. :)
Thanks for the references as well
There is not much use of bothering with the details, because you are quite powerless: you can't make ACR perform a different de-mosaicing, and DCRaw is no substitute for ACR.
I am not sure, but I think better exposure will eliminate or reduce the problem. This shot is almost three full stops from the right edge. On the other hand, you must not expose to the right without really needing that. Your camera compresses the raw data applying a lossy method. The pecularity of that method is, that the loss occurs in the highlights. Reducing the brightness in the raw processing can lead to "gaps" in the levels, which may cause posterization.
I neglected to explain, that the example I linked above is not from a D70. It was created by my Canon 40D, which does not exhibit this behaviour of the green pixels normally, it was the consequence of very strong flaring (I was testing the flaring characteristics of lenses). Therefor I can't tell with certainty, that higher exposure would eliminate or reduce the mazing.
I took another look at the raw file you posted. The difference between the greens in the row pairs seems to be excessive. The worst pair I saw was (155, 217). Note, that this is in the uncompressed range. If you have raw images from another D70s, pls post some, I would compare them. Perhaps this is not normal with that camera.
G Sch -
I may have access to another D70s. If so, I will post some raw files.
Also of note: I took about 500 pictures in the set that the uploaded raw file was from. I shot five different locations, and I believe only pictures from that particular setting (grey background, grey rug, softbox lights, etc.) exhibited the noise in question. I am scanning the rest of the set to see if I can find other examples of the noise in different settings, but I haven't found any yet.
I am not super familiar with the technical details of raw images - you suggest that there are multiple raw images which are combined?
Aha. You used flash on this shot? The shutter speed is 1/1600 in the exif in the file. You cannot flash sync this fast on the D70s. The max is 1/500s. Simply lower your shutter speed when you're doing flash and this thing will be gone. There is no reason to go this fast anyway, better to stay below the mechanical shutter anyway (1/250s). It is some sort of temporal interference I think between the electronic shutter and the flash.
> I believe only pictures from that particular setting (grey background, grey rug, softbox lights, etc.) exhibited the noise in question
Hold on. Did you use some filter as well? Was there some strong light from the side?
> you suggest that there are multiple raw images which are combined?
No. I am not sure what you mean, perhaps the two different greens?
The pixels sites have filters over them: green, red or blue. They are arranged in a regular pattern; in case of the D70 this pattern consists of two rows of two pixels each. The first row contains a blue and a green pixel, the second row starts with a green and then a red.
If you download this image and view it in at least 600%, then you see this arrangement. (This image is very greenish, mainly because half of the pixels is green, while only 1/4 of them is red and another quarter is blue. This is a non-demosaiced image.)
The values of green pixels, which are "diagonally neighbours" should be close to each other on the relative uniform areas, like the face, but they are not. You can verify that on this image as well if you increase the magnification even more, so that you can navigate over the pixels; however, the numerical differences you see here are much reduced from the original differences.
Anyway, I was referring to the green pixels in the alternating rows, once beside reds, once beside blues; they responded very differently.
Anything about that location that may suggest very strong radio interference? I have seen that exact pattern in shots taken in a studio that was just above a radio or TV station (can't remember if it was a TV or a radio station). The photographer had to move.
P.S. this means that I was wrong and that your camera is probably fine. This is just what happens if you use a way too fast (electronic) shutter with flash equipment. The photosites are not all exposed at the same time and you get these weird interference effects.
We will hear from Andrew if he used flash, but the girl's eyes indicate, that the lighting was strong without flash.
Anyway, I would be surprized if the shutter synchronization caused such an effect over the entire image, although who knows, what the electronics is doing.
Ramón's idea is not absurd either. The interference from some Canon lenses caused moire with a certain camera.
All of these shots were taken with the following lighting:
* a very large "softbox" style light from the right side of the picture
* a second smaller light (i believe) from the left side
* a set of flashes triggered from a Pocket Wizard on my camera
This was my first time shooting in this location, so I'm unsure about radio interference. However, shots taken at the four other setups in the room for the event showed no such patterns - this does lend some credibility to the flash being problematic?
>We will hear from Andrew if he used flash
The exif indicates that he fired a flash. And indeed it sounds like he did.
>this does lend some credibility to the flash being problematic?
Yeah, it is the flash. Make sure your shutterspeed never goes below 1/500s on that camera if you use flash. You might have to dial the flash intensity up if your other lights are very bright.
I'm relatively fresh to all this and just did a shoot with pretty much what sounds like the same settings as Andrew and exactly the same horrific outcome:
1/1600 sec shooting
single external strobe flash close to the subject
USB tethered connection shooting from my Mac with Nikon Camera Control Pro
RAW file workflow to Adobe lightroom
Running OSX Leopard
Wow. I shot all day with models over and was too stupid to notice they are stuck that way. I think I assumed it was some kind of RAW preview thing and that it would be ok once I turned the files into something editable.
Now I have a giant folder of the mosaiced images.
Anyone have a suggestion for the neatest cleanest way to remove without downscale and blurring?
How are you shooting strobes synced at a a 1600th of a sec on a d70?
like I say I really am fresh off the boat on this one so I'm not actually sure how to answer that question.
I basically eyeballed the light metering making sure the curves worked in at acceptable levels (not blown out anywhere etc) and the details were all coming up well. So yeah I'm pretty sure I screwed up what I should have used for ideal settings. My photographer friends are telling me I could have done it smarter but basically I guess now I have these photos I'm hoping there may be some software or photoshop filter combo solution out there for dealing with this problem.
I only asked because a d70 can only go as fast as a 500th flash sync,so anything faster and you will be getting partial frames.
At 1/1600 of a second the flash probably isn't contributing to the image at all and the extreme grain may be because the image is severely underexposed.
Sounds correct. i was thinking he would have problems from the flash in a portion of the frame,but maybe that is too quick to even register.