Saving for web

[Juan Conejero]

The direct link for these images is http://www.astral-imaging.com/M63-RGB-RC.html Now I know everyone has a prefernce for image processing but I wouldn't mind some feedback on the background levels and what appears to be most appropriate.

The first version is quite correct. Version a is too dark and version b is extremely dark. In deep-sky astrophotography you really want to show the faint stuff. Under normal daylight watching conditions, this is only possible if your mean background is relatively bright. For images dominated by free background areas like this one, I usually set the median values between 1.1 and 1.3, even if this implies leaving a small gap of unused space at the left side of the histogram.

I have three additional observations about your image. First, the image has a blue cast (look at the histogram peaks). The background should be neutral in my opinion, although the overall color balance seems quite correct. Second, the image has a "Generic PnP Monitor" ICC profile assigned. As I have noted in my previous post, this may cause problems for web deployment. Although this "PnP Monitor" profile is very similar to sRGB, you should convert the image to the sRGB color space and assign the sRGB profile when saving JPEG versions for the web. Last, I see objectionable JPEG compression artifacts; I'd use a higher JPEG quality.

[Juan Conejero]

I see objectionable JPEG compression artifacts

Wait, after closer inspection, this looks like aliasing artifacts. Have you rotated the image? If yes, which interpolation have you used?

[bitli]

Re Juan's remark 'the image has a "Generic PnP Monitor" ICC profile assigned.':

You should check the Edit / Color Management Setup configuration.

At least in my environment (Windows 7, calibrated monitor), the Default profile is set to the ICC profile of the monitor after an installation of PI and calibration of the monitor. This is wrong and should must be changed to the profile you want to work with (a wide gamut or sRGB). Only the profile of the screen must use the ICC profile of the screen (this is set correctly and cannot be changed on Windows).

I do not known when the settings is overridden, maybe in fresh install of PI or a reset of the settings or new calibration, but it is worth checking it after these operations. Naturally you can also set the profile specifically on each image you create, but sooner or later you will forget.

-- bitli

[Juan Conejero]

Aside from the browser rendering and colour space issues that Juan mentions, you also need to use a well-calibrated monitor to create your image in the first place. A colorimeter is required (e.g. one of the Spyder devices).

In my opinion, if you use a modern high-end monitor this is no longer necessary in practice. For example, I am writing this on a workstation with a DELL UltraSharp U3014 monitor. This monitor covers the complete sRGB space and a 99% of the Adobe RGB space. It comes factory pre-calibrated to ensure color fidelity with average DeltaE < 2 (basically, this is "perfect") and luminance uniformity between 97-102%. I work with this monitor set in Adobe RGB mode with the Adobe RGB (1998) color profile assigned in PixInsight (I use Linux) to produce lots of images, even images that are to be rendered in critical contexts (preprint). I don't use any hardware calibration and have no problems at all. It's a matter of watching the histogram and some statistical properties to keep everything under control.

[sreilly]

I just printed this topic out and will go over all the material and respond shortly. I did look at the referenced monitor and had several suggestions on using Ezio and NEC Spectraview monitors but find those extremely expensive for amatuer use. Now if I was generating an income.....



Thanks,

[Phil Leigh]

Aside from the browser rendering and colour space issues that Juan mentions, you also need to use a well-calibrated monitor to create your image in the first place. A colorimeter is required (e.g. one of the Spyder devices).

In my opinion, if you use a modern high-end monitor this is no longer necessary in practice. For example, I am writing this on a workstation with a DELL UltraSharp U3014 monitor. This monitor covers the complete sRGB space and a 99% of the Adobe RGB space. It comes factory pre-calibrated to ensure color fidelity with average DeltaE < 2 (basically, this is "perfect") and luminance uniformity between 97-102%. I work with this monitor set in Adobe RGB mode with the Adobe RGB (1998) color profile assigned in PixInsight (I use Linux) to produce lots of images, even images that are to be rendered in critical contexts (preprint). I don't use any hardware calibration and have no problems at all. It's a matter of watching the histogram and some statistical properties to keep everything under control.

Juan, I agree but such monitors are expensive - I can't afford to spend £1,000 on a monitor (especially as I have 3!!!)

[sreilly]

Juan Conejero replied

I have three additional observations about your image. First, the image has a blue cast (look at the histogram peaks). The background should be neutral in my opinion, although the overall color balance seems quite correct. Second, the image has a "Generic PnP Monitor" ICC profile assigned. As I have noted in my previous post, this may cause problems for web deployment. Although this "PnP Monitor" profile is very similar to sRGB, you should convert the image to the sRGB color space and assign the sRGB profile when saving JPEG versions for the web. Last, I see objectionable JPEG compression artifacts; I'd use a higher JPEG quality.

How would I go about correcting the blue cast? The process included background neutralization and color correction while in the linear stages. SCNR to remove the green was used. Do the same for blue?

The JPEG quality was set to 90%

[sreilly]

I see objectionable JPEG compression artifacts

Wait, after closer inspection, this looks like aliasing artifacts. Have you rotated the image? If yes, which interpolation have you used?

The image was not rotated. The only time I might change the orientation would be after the image is processed and then use the Geometry tool for presentation and save the result.

[sreilly]

Re Juan's remark 'the image has a "Generic PnP Monitor" ICC profile assigned.':

You should check the Edit / Color Management Setup configuration.

At least in my environment (Windows 7, calibrated monitor), the Default profile is set to the ICC profile of the monitor after an installation of PI and calibration of the monitor. This is wrong and should must be changed to the profile you want to work with (a wide gamut or sRGB). Only the profile of the screen must use the ICC profile of the screen (this is set correctly and cannot be changed on Windows).

I do not known when the settings is overridden, maybe in fresh install of PI or a reset of the settings or new calibration, but it is worth checking it after these operations. Naturally you can also set the profile specifically on each image you create, but sooner or later you will forget.

-- bitli

Looking at the Color Management and I changed the settings not to use the monitor profile as I had it set earlier. The attached screen shot shows how it is set now.

Juan, will these settings carry over after any version updates?

Thanks,

[sreilly]

One other issue, and probably the most important, is that I am probably posting to large an image considering my sky conditions. Being in the country at a reasonably dark site http://cleardarksky.com/lp/DgwdRdgeOBVAlp.html?Mn=dobsonian I have to contend with area neighbors with smokey wood fires in the winter and not so much outside light directly nearby but a correctional jail within say 35 miles that make use of sodium vapor lamps. It is apparent on cloudy nights as the glow raises up to maybe 20 degrees visually. From the observatory you can't see this as it has woods around the perimeter. I basically opened up about 1.5 acres out of the ten wooded acres when we built the house so any low horizons are blocked visually.

I can clearly see both winter and summer Milky Way on clear nights and my average seeing is 2-3 arc seconds as measured in my images. Best nights get 1+ arc seconds for periods of time. Coatings on the RC are about two years old and in very good shape. I avoid imaging over my house roof whenever possible to avoid heat currents. My image scale is .49 and .65 arc second/pixel at bin 1 for the ST-10XME and STL-11002 respectively. I usually post several image sizes to the website using 3,000, 2,000 and 1,600 pixel wide images. Maybe too large considering? Or maybe my processing is too aggressive? I don't usually crop the sides of my STL images even though I don't use a field flattener and the edge stars are distorted, maybe I should if it doesn't take away from the main object. Of course I could always revert back to imaging with the Tak FSQ-106 and do widefield imaging at a lower resolution. I think that puts me down to about 3 arc seconds/pixel.

I really appreciate all the information and feedback.

[sreilly]

After Juan mentioned the blue cast I opened the image and looked at the Histogram and then used the curves tool adjust just the blue channel. I think I resolved the issue and posted a separate page for the newly reprocessed image. The direct link is http://www.astral-imaging.com/M63-LRGB_reprocessed-RC.html . Hopefully this will correct the matter. Now the next issue will be correcting the monitor problem.

-Steve