Flame Nebula - 8 bit v 16 bit

[steppenwolf]

Sadly, I don't always practice what I preach but the poor weather recently has given me an opportunity to re-process one of my December images in a different way - the way I recommend!! This time, i re-processed the data maintaining its 16 bit attributes until the last moment. Although I normally recommend this if asked, I haven't been very good at keeping the images in 16 bit format when transferring them into PhotoShop but based on the following two images, I wish I had done so in the past!!

The original image processed in 8 bit format:-

The same data processed in 16 bit format:-

[barkis]

Yes, the difference is plain to see Steve. Now slap yourself on the wrist and say, "Must practice what I preach". And all will be well next time.

Lovely image too, both of them really, but No 2 is bestest.

Ron.

[centroid]

The difference between 'chalk and cheese' Steve

I could be wrong, but my guess would be that you'll be sticking with 16 bit from now on.

Dave

[Ant]

Blimey, I wouldn't have beleived that would have made that much difference...

Were the images processed in the same way (except for the 16bit part)?

Ant

[steppenwolf]

Were the images processed in the same way (except for the 16bit part)?

Yes they were except, of course, I had a LOT more latitude to play with the stretch and it was much less 'destructive'. When I processed the original image, I concentrated on getting a nice 'red flame' and having got that, I worked on the smoothness by blending in a Neat Image filtered overlay. I did exactly the same this time and aimed at getting the red of the flame to a similar intensity. The image, before being filtered by Neat Image, was at this stage already better than the original 8 bit version so the filtering gave a nicer layer to blend with the base image.

I think that the main reason that the 16 bit image is (to my eye at least) better is that the 8 bit original is that stretching to get the red portions right did not destroy other parts of the image (like the Star Alnitak) because I just had so much more control over the process as there was just such a lot of available 'stretch capacity'.

[LaVey]

Can you explain the 8 / 16 bit thingy here.. is it bmp to jpeg conversion?

I use a meade LPI and think I remember someone saying that these only produce an 8 bit image anyway?

[radioactive]

bit of difference between the two images there well at least 8 bits for sure!

seriously though it really does show up in the end don't it

BTW 2nd one is best in case anyones wondering :insects1:

[Trevor]

Hi Steve

That’s some image difference: wonder what your Crab Neb would look like, that’s if you still have the data.

Trev.

[steppenwolf]

Can you explain the 8 / 16 bit thingy here.. is it bmp to jpeg conversion?

It is not directly to with conversion to .jpg BUT a JPEG image and for that matter a .BMP can only be in 8 bit format because of their architecture whereas, for example, a TIF file or a FITS file can be represented in 16 bit format. The number of bits describes the number of colours and their levels available in the image. An 8 bit image can only display 256 levels per pixel and 256 shades each of Red, Green and Blue ( 2 X 2 X 2 X 2 X 2 X 2 X 2 = 256) whereas a 16 bit image can display 65,536 levels per pixel and 65,536 shades each of Red, Green and Blue!!!

[cygnusx1]

So let me get this straight,

After aligning and combining the lights, bring them into PS, but dont save as TIFF/FITS until you have completed all post processing??

What about if you need to post-process in in Maxim or PS Elements?

I'm probably right off track here

[steppenwolf]

Hi Kevin,

To bring them into Photoshop you DO save them as a TIF (not a .BMP as I used to do!) but tick the 16 bit option. This will keep the data in 16 bit format. Photoshop will allow you to process the images but some options will be greyed out including CROP, REPLACE COLOUR, DESATURATE, SELECTIVE COLOUR, THRESHOLD, POSTERIZE and VARIATIONS as will some of the filters be greyed out too (for whatever reason, these functions will only work on 8 bit images). The most important functions, Levels and Curves work fine. It is only after the main work has been done that you may have to resort to converting to 8 bit to carry out the other functions.

[robindurant]

Hi Steve - You have brought up a subject I have been working on lately. There is another factor that I would like considered and that is the fact that most of these images we view are from or on a Computer screen. I process my images on a PC due to the fact that my PC has CS2 on it. However when satisfied with the result I transfer it to my MAC and find that on MAC the image are a poorer quality. I then re process them on Elements 4 and obtain much better colours than the originals. I have also noticed that often when posting an image on SGL the result is poorer that the original I posted. Maybe saving images in 16 bit would prevent this image posting deterioration.

Robin

[steppenwolf]

Maybe saving images in 16 bit would prevent this image posting deterioration.

Robin

Hi Robin,

The problem with this is to do with image size and internet bandwidth. Image files on the internet tend to be stored as JPEGS because they can be compressed in size so that (a) they take up less space and ( because they are smaller they transfer across the net quicker. This raises two issues:-

1. JPEG files can only be saved as 8 bit files

2. The compression algorithms of the JPEG standard are 'lossy' - i.e. permanent 'damage' is done to the data to make it take up less space and this 'damage' manifests itself as artifacts in the end image. The higher the compression, the more lossy the image becomes.

It is also important to note that every time a JPEG is opened, manipulated and then re-saved, accumulated damage occurs so any image should only be finally saved as a JPEG when every other step of the processing had been completed. When you transfer your images from your PC to your MAC you should absolutely avoid JPEGS at this stage.

Another issue between the MAC and the PC *MAY* be 'colour space'. In this wonderful world of computers there is more than one colour 'pallet' that can be used to represent an image - RGB, sRGB, Adobe RGB, CMYK, HSL and others - it is important to ensure that the same colour space is used on both your systems to ensure similar viewing results although you still need to bear in mind that different monitors will display your images differently as well!

[MartinB]

No comparison Steve, the 16 bit has much more depth to it. A fantastic demonstration and thanks for sharing - a very useful post.

I think this 16 bit and file format stuff is really important so I'll stick in my twopennies worth for those who might be a bit confused.

16 bit allows 2 to the power 16 brightness levels >65000 whereas with 8 bit it's 2 to the power 8 or 256.

Now a monitor can only display 256 levels so photoshop only displays an 8 bit histogram. However, it recognises that 16 bit image and all it's levels. So within each 8 bit level are 256 further levels. Most astro images are way over to the left of the histogram. When we stretch the image we are pulling those levels over to the right and broadening the histogram. All of a sudden these hidden levels come pouring out and the resolution of the displayed image just keeps on increasing.

Not all 16 bit formats are the same though - FITS allow a floating decimal point which means that when mathematical functions are performed on the pixel values e.g when stretching the histogram the resulting change is more accurate and less noisy than with a TIFF image which only allows 16bit integer values.

These are my rules for handling files. Work with FITS for as long as you can. PS can't handle FITS but newer versions can with an appropriate plug in such as FITS liberator or FITS plus (which handles colour fits unlike liberator). Always save FITS in a floating point format if possilble.

If you have an older version of PS then you will need to use tiffs. If this is the case, strech your floating point FITS using DDP before saving as a TIFF - this minimizes the integer data loss.

Only convert to an 8 bit jpeg when you have finished processing

[steppenwolf]

Thanks for the additional information Martin. I have FITS liberator for Photoshop but it is a real pain to use for colour images as you have to load it several time and extract a different 'plane' each time and then re-combine the channels to make an RGB image but I have not tried it to determine if the 16 bit floating point aspect is maintained - I'll give it go and see!

One of the most striking differences in the two images above is the treatment of Alnitak. This is my star of choice for 1 star alignments when taking images within Orion and I have often noticed the little star next to it in my focussing subs. This star was completely burnt out in the 8 bit stretch but clearly visible in the 16 bit stretch.

I love this hobby - always something going on to keep the brain active!

[MartinB]

Steve, you need a different FITS plug in to liberator for colour. Centroid put me onto it, think it's called FITS Plus. I'll check it out. You have to pay but it's not a huge amount.

[steppenwolf]

Thanks, Martin, I'd be very interested in it.