Hello,

Is it possible to generate a 3d volume with depth from an incoming image? Request is to convert a figure from an image into a cloud like structure, and Im trying to figure out an intuitive way to do this.

Currently Ive just manged to take an image mapped to the uv coordinates of a piece, scatter points based on the color, and then create a volume from that, but that is all i have so for.

My thought was to perhaps deform the geo based on the luminance values of the image so as to give it more depth, but overall Im not really sure how to approach this.

Ive attached a pic and the current scene file.

Any tips would be appreciated. Thank you!

not sure if houdini has functionality for that built in somewhere, but there is a lot of software out there that will generate a depth map from an image.
http://depthy.me/#/ [depthy.me] i hope it's okay to post links here.
i'd use that to displace a tessellated plane, cap the back and turn it into a volume.

Interesting assignment! My initial though is that you will need a model of the mandril's head. Either model it, or get a geometry from TurboSquid or similar. Then using the image of the mandril , adjust the existing geometry to match as close as you are comfortable with.
When geometry is done, give color (Cd) attribute to the geometry using AttributeFromMap SOP. You will need to use front projection as UV.
Now using the geometry with colored points , generate the point cloud using scatter tools, or particle DOPs.
Next, using DOP I/O SOP, import the DOP POP pointcloud into SOP context , in the same Geometry Object where your mandril head geo resides.
Next use AttributeTransfer SOP pipe point, primitive, and UV attributes from source geometry to the point cloud.
Cap the work with null SOP named nullOUT. That location to render ROP.

On to rendering!
At object level set Render parameters to use Render As Points-Spheres with small radius like .001 or as desired.

Hey thanks for the tips both of you. We have actually killed the project. The image that they were trying to use did not relay itself well to the project at hand anyway–it was an image of a guy against a comped sky, the structure of the image was not helpful. If the the project had not been killed though, we would have definitely gone the route of modeling a decent replicated mesh and projecting color data to drive density. As an aside–if this was for a different sort of project and there was more complexity to the scene itself, there is some neat software out there for generating depth maps from images–might be a useful thing to be aware of. Apologies for the delayed response, but thanks for the tips!
Best,
JK