Conifer needle surface area minus shadows

[Tree-Doc]

I have about 250 scans of conifer tree needles and need to measure their surface area. I've been using Threshold Colour to adjust the HSB to remove part of the shadow, but I notice that depending on the exact threshold numbers I use, my resulting area calculation will vary by 25% or more (not good).

Is there a better way to do this to achieve more accurate results?

My methods: Adjust HSB to remove as much of the shadow as I can without removing green needle.
Converting to binary.
Despeckling the noise.
Process>Binary>Close to capture the white center streaks in the needles.
Filling the Holes.
And analyzing particles

Here is an example of the images I'm working with. I know the scans aren't ideal, but capturing better images in not an option.