Re: [APML] Fw: images

["Carlos Milovic F." <cmilovic@yahoo.es>]

>I'll repeat what I said before. 6 bits per channel is 262,144 possible
>colors. You won't see the difference between that many and 16.7 million >possible colors in any one image.

Perhaps not, but I do if we are talking of 16bit and 24bit (3 channels)... and I'm sure I can see the steps in a gray 6bits gradient. As I said, we need at least a little over 7bits to see a continuom

>> For all practical ways a photograph of a continuom gray gradient
>> indeed looks continuos... and the human eye needs a little more of
>> 7bits to see that.
>That's probably true in viewing the real world but not a monitor or a
>print.

? :-S

I don't understand what you are saying...

>> Anyway, due the human eye's limitations, the scanning process, the
>> monitos, etc. I think that you have a strong bias in your medition...
>> that makes appear the film's "bit deph" worse than it is.
>Perhaps, but compare my 10 bit image example to any astrophoto.

Ok. Here we are talking about diferent matters. Astrophotos make use of other zones in the dinamical range, and usually the information we want (don't think on the stars, they cover the whole Drange) is located in very few levels. I recognice that indeed if we count how many levels we use from the original scan it could be 2^6 per channel (in a 16bits file, scanned at 10bits). Anyway, it doens't mean that film has a bit deph of 6bits.

>> Just a final thought... if film can handle only 5-6 bits, why do we
>> have 16bits scanners? Why not just 8bits ones?
>Marketing! Sorry, I couldn't help but be cynical. ;-) Really though, my
>old SprintScan 35ES which scans in 10 bits and outputs in 8 bit is
>quite adequate except the Dmax it can handle isn't high enough.

I prefer to talk about bit deph just as a matter of precition... Imagine that we have nice curves controls in the scanning process, and that we can scan the original and obtain the directly "final" result. How many grays do we have? Think that the colors are represented in the 0 - 1 range, with the 16 bits as precition. I think that we'll have more than 64 grays. Perhaps not 2^16, becouse the original data isn't stored strictically as a continuom in the film, but indeed more than 64.

>> PS: Photoshop is a nasty program... as Vicent said, it destroys the
>> data. I don't trust the ways it works. Just make a gaussian blur,
>> you'll see steps in the gradients.
>I tried it. I made a continuous gradient from black to white and
>performed a guassian blur. No steps here. It is very smooth at any
>pixel setting of the blur.

Use large radii to blur a astronomical image, just as if you were creating an AV mask with that procedure.

>> Also there is a problem with the monitor's capabilities of displaying
>> the data. Even if you have a continuos (at >16bits) gradient you'll
>> see some steps.
>No monitor I know of is capable of displaying 16 bit grayscale. As I
>said above, I have no steps in a continuous gradient. Maybe you have a
>hardware problem?

I know that no monitor displays more than 8bits per channel. I have a color gradient beetween few levels at 32bits per channel (generated throught a function). I can see some steps in the gradient, but if read the values there are none... And it isn't a hardware problem (video card). Just the monitor isn't capable of displaying the whole range of colors.

 

 

Regards,

 

Carlos

---

_______________________________________________
Astro-Photo mailing list
Astro-Photo@seds.org
http://seds.org/mailman/listinfo/astro-photo