A RGB-D image is simply a combination of a RGB image and its corresponding depth image. A depth image is an image channel in which each pixel relates to a distance between the image plane and the corresponding object in the RGB image.

You can use Kinect to capture such RGB-D images. If the use of Kinect-like hardware is not available in your task, you may need to estimate the depth from images of the same scene taken from multiple cameras, coming down to a computer vision problem. Another solution is to collect traning data containing RGB-D images and to use machine learning techniques.

This may help: https://www.cs.washington.edu/research-projects/robotics/rgbd-mapping

But nevertheless, there are plenty of methods to retrieve depth from images. The selection of the proper setup and method depends of course on your specific application (which I do not know). For a specific answer you might shortly outline the application you have in mind.

Here are some google key words which may help:

shape-from-stereo (there are also techniques using multiple images instead of only two images)

shape-from-shading

depth-from-focus (microscopy)

...

Here you find a nice overview of some methods with a critical discussion:

http://hci.iwr.uni-heidelberg.de/publications/dip/1992/Jaehne_SPIEProc1992.pdf

here is another approach:

http://www.automationtechnology.de/cms/en/3d-imaging/3d-technology/sheet-of-light-method.html

Thanks to all for nice help.

It refers to the number of chanesl that correspond to the color model