Banding in night sky

[mpd]

<p>I was taking pics of Boat House Row in Philadelphia last night. The below pic is the result of a 10 Exposure HDR in Photomatrix. It was shot at f 8, from 1/10 to 60 sec. I opened the result in CS3 and did a slight curves adjustment to correct the color balance. It was done at a 16 bit color depth.<br>

There is banding in the sky. How did I create it? How do I get rid of it? I didn't want to do any more PP until I figured this out.<br>

<a href="http://farm4.static.flickr.com/3660/3338164296_2bd9ed1c19_b.jpg">http://farm4.static.flickr.com/3660/3338164296_2bd9ed1c19_b.jpg</a><br>

Sorry for the large pic size but I wanted the banding to be visible.<br>

thanks,<br>

pat</p>

[howard_m]

<p>what was your original capture mode? raw/jpeg? if raw, what sw did the conversion?<br>

what color space are you using?<br>

on a different note, IMO, this shot really doesn't require 10 shot HDR (maybe more like 3, maybe 5). I suppose it's possible that as photomatix works, it's quantizing the low levels and that's the source of posterization.<br>

try a 3 shot HDR (or enfuse/hugin).</p>

[robert_brake1]

<p>I don't know the answer but I would be really happy if you keep us informed as you work on the problem. I have found that Adobe's tone-mapping from HDR in PSCS is really bad when it comes to smooth gradients: banding, posterization, etc. I looked at the Photomatix and it was better but not the end all. I agree with Howard that other software products may be the key. Finding a better solution than masking, blurring, etc., would be nice but gradients seem to simply be a problem. I typically shoot simple -2, 0, -2 images when looking at the banding problem and, although they may be better than high image counts, they don't seem to be the complete solution.</p>

[howard_m]

<p>it'd certainly be simple enough to mask and do a 1px gauss. blur on the sky to eliminate it. there's nothing up there.</p>

[roger_smith4]

<p>Consider using 16 bit TIFFs to generate your HDR file if you were using jpegs. </p>

<p>One workaround for this file would be to open it and a layer where the sky is about as dark as this and using a layer mask, brush in the sky from the good image.</p>

[howard_m]

<p>start with raw, as many bits as the camera has. I think the 8 bit intrinsic nature of jpeg (if thats your starting point) coupled w/ 16 bit processing and probably too much working space (proPhoto may be working against you if you're using it)</p>

[mpd]

<p>Update. These were all shot in Raw. I let Photomatrix operate on the raw files. I dont' see the banding in Photomatrix when I view the image full size. <br>

Here is something unusual. When I open the Photomatrix processed .tif file in CS3, the bit depth is listed as (RGB16*). I have never seen the asterixsk before. When I convert to 8 bit, it remains(RGB8*). However if I convert from (RGB16*) to (LAB16) and back to (RGB16), the asterisk disappears. Could there be some difference in the 16 bit .tif output by photomatrix and what CS3 expects to see as a 16 bit .tif?<br>

I just redid the conversion, same result. Now I am trying DPP to convert to tif. I will then retry in Photomatrix and in Picturenaut. I will post those results later this afternoon.<br>

Howard, I have a new toy, a Panocamera modified Nintendo DS controller. I got it working Fri night and my trip into Philly was a test run. I wanted to see how shooting lots of brackets compared to just shooting the -2,0,+2 that is available on my Canon 1D Mark II. Last summer the -2,0,+2 was inadequate in San Chappelle. </p>

 

[howard_m]

<p>I wonder if the asterisk means that there is/was a profile mis-match between what PM output and the working space. I dont think LAB actually has a profile per-se so the roundtrip conversion nullifies the mismatch. Not sure why you're doing the LAB round-trip or the 8 bit conversion.</p>

[john_latimer1]

<p>Pat<br>

You can do a very simple process of adding a very small amount of grain/noise to the affected gradient area. This is much better than blurring which in many cases actually makes the problem worse. If you use the add noise filter I often use the monochrome option. It does not take much to work. The result of noise is that it disrupts the pattern of the image without doing much more and this disruption is enough to break the band but keep the overall gradient. While not the perfect fix it can work.<br>

In the near future I will have an article out on an HDR alternate inside of CS3/4 if you have the extended version. Email me and I will send the article to you when I am done.<br>

john@ejarts.com</p>

[mpd]

<p>I processed the Raw files into Tifs using DPP. I open the Tifs and processed the HDR using both Photomatrix and Picturenaut. I saved the results and both produced images that showed banding in CS3.<br>

Then I opened the HDR Tif in Photomatrix and no banding was visible. I opened the same file in CS3 and banding was visible. I put the images next to each other and saved the result. Does this offer any clues?<br>

http://farm4.static.flickr.com/3333/3338743654_3dc12b4d12_o.jpg<br>

I think I will post in the Photomatrix forum to see what they have to say.<br>

Thanks for your ideas.<br>

pat<br>

PS John, email is on the way.</p>

[Tim_Lookingbill]

<p>My only guess to the cause of the banding harkens back to the fact that the linear characteristics of the sensor data distributes the brightest parts of the image captured (the first stop) to half of the entire high bit tone scale starting somewhere around 128RGB on up to 255 in terms of a gamma encoded scaling (as opposed to linear sensor scaling).</p>

<p>Since your sky has mapped to a gamma encoded space close to 60 RGB, my guess is that your dark sky didn't fall within that first stop of brightness to provide enough of the tonal steps to withstand a curve edit in CS3. Even though its 16bits, its interpolated 16bits from a linear sensor tonal distribution system at the time of capture.</p>

<p>I get the same posterization editing shadow detail shooting at high ISO in somewhat low light scenes even though I've Exposed to the Right which results in an overexposed shot. No matter how over exposed I made the shot look I couldn't fool the sensor into kicking that shadow detail up the tone scale (the first stop of brightness) so it is rendered by higher amounts of tonal steps. The best I got was more noise in these areas and that may also be another cause to your banding issue. Varying levels of noise within each bracketed exposure blending together in Photomatic may have caused a kind of overlaid moire/conflicting pattern in these areas.</p>

<p>Just my guess because I'm not familiar with how Photomatix treats linear sensor data, but then I'm not that familiar with how any piece of software be it incamera or third party does either.</p>

[Tim_Lookingbill]

<p>M Patrick,</p>

<p>Nix what I just said above. I just saw your screencapture and the banding is far more prominent than in your first posted image. I don't see that kind of banding on my calibrated iMac of your first linked version.</p>

<p>My guess is that you have a bad display profile. Recalibrate using a hardware calibrator or load a canned profile like sRGB as your system profile and see if there is a difference in CS3. Also make sure that CS3 is using or recognizing your system profile as chosen by going into Color Settings and click and scroll within the RGB space dropdown menu to where it says MonitorRGB-XXX. Make sure the name after "MonitorRGB" says the chosen system profile.</p>

[mpd]

<p><b>Technology sometimes get in the way.</b><br>

I think I may have stumbled upon the problem while following Tim L's advice. But before I explain, let me tell ya a little story about how technology sometimes gets in the way. Wed of last week I was getting ready to leave the house around 3 PM to run some errands. I had been working from home all day and really just needed a mindless break. For some reason I logged into facebook to see what was going on. The top post to me was a plea for help from my son's finace. Seems my son has a fever in excess of 102, was in a lot of pain and they felt he should go see a doctor. While my first thought should have been,"You're right. Let me find you one", it wasn't. My first thought was, "had I not happened to check my account this could have gone unnoticed for quite some time". Yet clearly it needed attention now. Had they become so enamored with IM's and posting on facebook that they had forgotten how to use the telephone?<br>

I think a similar usage of technology is at the root of my banding problem. I have 4 computers in my office and countless virtual machines(don't ask!). Yet I only have one monitor and keyboard. I have a KVM(keyboard, video and mouse) switch but I stopped using it because as soon as I switch physical computers with the switch, I loose the scroll capability of my mouse. I have started to use Windows remote desktop connection to operate the physical computers. While I was trying to see what color space was being used on my photoediting computer, I noticed the bit depth was only 16 and not 32. I checked my remote desktop setting and I was asking for 32 bit support but wasn't getting it. <br>

Conclusion, when I am looking at the sky through a remote desktop connection, the sky is banded. When I switch my monitor to that computer directly without closing the pic or CS3, the banding is not visible.<br>

I out teched myself. Sorry for all the confussion. But thank you all for your responses. <br>

pat</p>

[patricklavoie]

<p>add a bit of monochromatic noise like 2-3 all over or with a mask..problem solve!</p>

<p>next? : )</p>

<p> </p>

[philfx]

<p>I dont see a strong banding. There is a slight bit that a blur can easily solve. I do see a stronger "clitch" top right of the sky, and a small blurp left of the tree at far left edge. Hmm, do you see strong banding? Also, do you see it in a hi res print? I would try that.<br>

Cool image</p>