If you use Mask_All in your request(), input0().request(x,y,r,t,Mask_all,count)
then it will read all of the available channels from the input. If the depth.z
channel exists in your input then using Mask_All in your request function will
load the depth data.
In the engine call “channels” is equal to your set_out_channels() in your
validate() function. If you set_out_channels(Mask_RGB) then the engine will use
Mask_RGB for channels. Then when you use foreach(z,channels) the engine will
only cycle through RGB even though there may be additional channels available
for access. So you can access the depth information, but when using
foreach(z,channels) the engine will not include any modifications to the depth
information.
If you still want to pass through the depth information (ie
set_out_channels(Mask_All)) you can always include a check in the foreach loop,
something like if (z == Chan_Z){ [copy input to output] }
In your engine you would create a row and fill it with the depth information:
Row d(x,r)
d.get(input0(),y,x,r,Mask_Z) //fills row with Mask_Z (depth.Z channel)
You can then get the depth value for each pixel X with:
float depth = d[Chan_Z][x]
Once you have the depth value you can use it as your tile search size in v_pass
(obviously you’ll have to modify v_pass function slightly) to be equal to the
depth for each pixel. It might help to create a full frame tile and pass it to
the v_pass function as well so that you only create one big tile for all pixels
as opposed to one small tile for every pixel. Then you can use the same depth
value on your x pass. If you want a separable x and y size you’ll need to set
up some sort of depth scalar value for each axis.
Disclaimer: BTW I haven’t compiled any of this so no guarantees on the syntax
being exactly correct, but hopefully you’ll get the idea.
From: [email protected]
[mailto:[email protected]] On Behalf Of Benoit Delaunay
Sent: Friday, October 24, 2014 10:51 AM
To: [email protected]
Subject: [Nuke-dev] Dilate a channel according to another channel's values.
Nukers, hello !
What I am trying to achieve is a node that would:
- read the Z channel
- dilate every channel according to the Z values
the exact same dilation would be applied to every channel.
What I tried:
LevelSet then iDilate, not perfectly accurate and too slow.
VectorGenerator on a 2 frames animation of the original z / dilated z,
not accurate enough either.
I tried learning C++ and the api but these _request, _get_vpass and
Row are a bit obscure to me...
I started with the Dilate.cpp source, which has these
foreach (z, channels) {
const float* FROM = in[z];
float* TO = out.writable(z);
float v = FROM[x + h_size];
I added this before this loop
const float* FROMz = in[channel];
float* TOz = out.writable(channel);
float vz = FROMz[x + s_size];
'channel' being the depth.z channel of the same input.
Then obviously I have to remove channel from channels
and try to delay computing TOz until I had dilated every other channel
if (z == c1.last())
First problem: I don't know exactly how to use this z channel and add it
to the 'requested rows', using it's values without actually altering them.
Maybe I could store the values somewhere else ?
I need the original values for each channel but I need the altered values
because the dilate.cpp modifies all these TO values twice
Second problem : a while loop prevents me from just using this .last() check.
I tried to remove this foreach (z, channels) loop by defining separate
variables for the r, g and b channels and now it crashes instantly.
Third problem : Am I even on the right track ?
Mayday !
Cheers
Benoit
________________________________
******************************************************************************************
This email and any files transmitted with it are confidential and intended
solely for the use of the individual or entity to whom they are addressed and
may contain confidential and privileged information. No one else is authorized
to distribute, forward, print, copy or act upon any information contained in
this email. If you are not the intended recipient, please contact the sender by
reply email and destroy all copies of the original message.
*******************************************************************************************
________________________________
******************************************************************************************
This email and any files transmitted with it are confidential and intended
solely for the use of the individual or entity to whom they are addressed and
may contain confidential and privileged information. No one else is authorized
to distribute, forward, print, copy or act upon any information contained in
this email. If you are not the intended recipient, please contact the sender by
reply email and destroy all copies of the original message.
*******************************************************************************************
_______________________________________________
Nuke-dev mailing list
[email protected], http://forums.thefoundry.co.uk/
http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-dev
