Question for PS gurus

[Jim Sanderson]

Is there an advantage of using smart objects to make lab adjustments to RGB images rather than just converting to lab and back to
RGB.

Jim

[PeterAit]

Is there an advantage of using smart objects to make lab adjustments to RGB images rather than just converting to lab and back to
RGB.

Jim

You'll be converting to LAB and back to RGB regardless of whether you use a smart object. There's no advantage to using a SO for this, as far as I know. THe only thing that SOs give you is making all edits reversible. But, there are enough things you can't do with SOs that I have given up on them.

[Mark D Segal]

The first question you need to ask yourself is why do what are doing in L*a*b*. There's almost no need to work in this colour space at all, and by avoiding it you minimize the data loss associated with the conversions, as well as the non-reversibility of the workflow involving the L*a*b* elements, so unless there is some very special reason to use it, I would recommend generally avoiding it.

[Jim Sanderson]

Thanks for the responses.  I appreciate your feedback.  I don't completely agree with MarkDS with regard to the use of LAB though.  I use LAB often for a variety of different reasons including channel blends, writing curves, selections based on color, use of the "blend if" functions and retouching.  I've been using the smart objects so that I can make adjustments in LAB or CYMK and then be able to go back into the smart object, if need be, to change adjustment layers and settings where in the past I could not do so by straight conversion without the use of SOs.  I think my question, although a bit vague, was whether because of the non destructive nature of SOs, do I lose image data at all by using SOs?  

You guys gave me some food for thought and it has been helpful.

P.S.  I looked at your  respective websites and read a couple of articles from Mark's and found them (both sites) to be informative and interesting.  Thanks again,

Jim

[JeffKohn]

The first question you need to ask yourself is why do what are doing in L*a*b*. There's almost no need to work in this colour space at all, and by avoiding it you minimize the data loss associated with the conversions, as well as the non-reversibility of the workflow involving the L*a*b* elements, so unless there is some very special reason to use it, I would recommend generally avoiding it.

I disagree. a/b curves are far more powerful than the Vibrance adjustment, and L* curves can also be useful. LAB is also useful for color corrections (sometimes getting the white balance right isn't enough, and you need to do localized corrections with masks or blend-ifs). The a/b channels can also sometimes be very useful for selections. The loss of data from a LAB round-trip is a complete non-issue if working with 16-bit files (as you should be anyway).

[digitaldog]

The first question you need to ask yourself is why do what are doing in L*a*b*. There's almost no need to work in this colour space at all, and by avoiding it you minimize the data loss associated with the conversions, as well as the non-reversibility of the workflow involving the L*a*b* elements, so unless there is some very special reason to use it, I would recommend generally avoiding it.

Yup, do that kind of work in the Raw processing pipeline.

[Jim Sanderson]

Thanks for the responses.  Appreciate them very much.  However, does anybody have an answer to the "posited" question as refined in my second post?

Jim

[robertjm]

I think my question, although a bit vague, was whether because of the non destructive nature of SOs, do I lose image data at all by using SOs?  

Jim

My gut feeling says 'lunches=free? NOT'  
I'd say that you would still incur the data loss of the double conversion, but you'll only see it later in the process when you generate output

[JeffKohn]

Thanks for the responses.  Appreciate them very much.  However, does anybody have an answer to the "posited" question as refined in my second post?

ACR renders RGB data. So if you move to LAB with the SO intact, whenever you edit the SO it's going to re-render to RGB and silently convert to LAB, I don't really see how it could do anything else. But I don't think you have to worry about data loss as long as you're working in 16-bit. The amount of data lost by RGB->LAB conversion is inconsequential for 16-bit files, and indeed less data is lost by that conversion than by most of the edits you would make in Photoshop yourself. If you're wondering whether this data loss accumulates each time you edit the SO, I think the answer is no; the SO re-renders starting from the RAW file each time, so it's not like you're accumulating multiple round-trip colorspace conversions.

[Mark D Segal]

My understanding is that most of the data loss happens on the first round-trip, and while it occurs thereafer too, it is not as much. There is a whole discussiom thread about the L*a*b* issue here:

Using L*a*b*. If a conversion of an 8 bit image loses something like 20~30 levels in the first round trip, that is a sizeable chunk of a 255 level image. It would be perhaps insignificant in a 15+1 bit image, unless the level loss were relative rather than absolute.