An actor has cg elements sticking out all over his head. The elements near the back of the head need to defocus to give the dof effect.

When using a depth map as a matte to blur the elements it doesnt seem to work well as it just blurs inside the area of the element leaving hard edges. If the depth map is blurred it kinda of works but its still weird..

Any tips or ideas on how to more effectively use the depth maps in comp to produce the dof for these.

ps. This is being comped in shake.

Using them strictly as mattes? That will only get you soft insides, as you've discovered. It's the same approach as with Houdini, just some persnickety Shake stuff you need to do. I use depth maps rendered as OEXR and use them with ZBlur nodes, works great. Probably more suited to a Shake forum , but you'll need to feed it to a reorder with “rgbal”, then a copy node and copy over that fourth channel only as “z”. Shake has a limited ability to handle different channels - you're forced to only work within the rgbaz paradigm. The docs get into that.

'Course Houdini needs no such acrobatics.

Cheers,

J.C.