Image Recognition and 3D Rendering

How difficult it would be to make an image of an object (in this case, a predefined object) and develop an algorithm to cut out only this object from a photograph with a background of varying complexity.

In addition, a photo object (for example, a house, a car, a dog, but always of the same type) must be converted to 3D rendering. I know that 3D rendering engines are available (at a price, for free or with some offer), but for this to work, an object (object) will need to be measured in various ways - for example, if it’s a person, we need to measure the height, curvature of the shoulder, face radius, length of each finger, etc.

What would be the feasibility of solving this problem? Does anyone know of good links related to this area of ​​research? I have seen open source solutions for this problem that leave me wondering about the simplicity of measuring an object while tracing around it to crop it.

thanks


Essentially, I want to take a 2d image (typical image: it’s easier than a complex photo containing several objects, etc.)

http://benmartin3d.com/WIP/Project1/image1small.jpg,

But effectively I want to turn this into a three-dimensional image, so not what I want to do is to create a 3d rendering / modeling engine?

In addition, this link that I provided is included in 3ds max with several sets of properties and is being rendered.

+5
source share
4

, , .

  • (.. )
  • 3D ( 3d )
  • ( , - )

( ). , . :

(Wikipedia)

( 3D-)

+4

, , , 3D- 2D-. , 3D- 2D-, "Shape from Shading" Google , (, , ) .

Rob

+1

. , . , ( , ;).

googling "shape-from-shading". . (BRDF ), , (, ), .

+1
source

Assuming this would be extremely difficult, especially with a single image of an object. The rasterizer must guess the depth and distance of objects.

What you are describing is very similar to Microsoft PhotoSynth .

0
source

All Articles