Yosemite B&W conversion -- can it be better?

[dancorder]

This is my first proper attempt at a B&W conversion (whipping post thread that started this here (http://www.dgrin.com/showthread.php?t=59371)) Given the original colour pictures below what could I have done better?

http://dancorder.smugmug.com/photos/145333645-L.jpg

What I did was take the green channel from each of the two cropped images below, play with the curves and used a mask that faded linearly from opaque to transparent accross the horizon to merge the two together (ie apply an ND grad filter to the original). I picked the green channel for the sky because it looked the most interesting and for the ground because it gave the least contrast between the mountains and the trees which meant that I could have more contrast over the mountains (in the other channels the trees were almost black with the mountains light grey)

Below are (in order) the original picture with minimal RAW processing, plus the two (cropped) RAW exposures I merged to make the image above. Looking at the sky exposure again I think I should have pushed that further in the RAW converter which might have given a bit less noise in the clouds.

http://dancorder.smugmug.com/photos/145458698-L.jpg

http://dancorder.smugmug.com/photos/145458681-L.jpg

http://dancorder.smugmug.com/photos/145458660-L.jpg

[dancorder]

I realise that my original post wasn't that clear about what I was asking, so given that I'm new to the world of B&W conversions what could I do to make my mine better? (Or have I lucked out on the first attempt :wink)

[rutt]

It does look pretty good, but maybe you can make it even better.

Take a good look at the red channel for the sky. The common wisdom is that the red channel is the best starting point for skies. It's relatively easy to blend. It will be the lightest part of the image and also the only part with much blue.

A really simple thing might be:

Duplicate layer
Apply image green to the composite channel of the layer in normal mode
Apply image red from the background layer to the composite channel of the duplicate in darken mode. Maybe play with the opacity.

You can also do exactly this with two separate layers, one for red and one for blue which will give you greater control of the blend.

[dancorder]

Thanks rutt,

I went back and had another look at the red channel and it did seem better than the green so I had another go. This time I also upped the contrast on the sky and the ground which made them harder to blend without it being obvious. I then had the idea of using a gradient fill on the mask but over a larger area. I then used the curves on the mask to change it from a straight gradient to one that rolled off gradually which I think helps hide the joins.

http://dancorder.smugmug.com/photos/149639785-L.jpg


A really simple thing might be:
Duplicate layer
Apply image green to the composite channel of the layer in normal mode
Apply image red from the background layer to the composite channel of the duplicate in darken mode. Maybe play with the opacity.You can also do exactly this with two separate layers, one for red and one for blue which will give you greater control of the blend.
If I had understood that I would have tried it :wink. Which layer are you starting with and what is the composite channel? I'm using the GIMP (and I'm no expert!) so I guess the composite channel might be called something else. What are the above steps going to do the picture? If I understood how they were supposed to affect it I might be able to work out an equivalent.