what's the different of 2D depth image with RGB image?

[xiaochongs]

Hi,

I am researching the 3D Face Identification. after reading the paper, I have some questions about the way:

1. represent a 3D facial point cloud with an orthogonally projected 2D depth map. As you know I can get a 2D image from rgb camera. what's the difference between the projected 2D depth map and 2D RGB image? I think a 3D object is same with 2D after with an orthogonally projected.

Thanks. Best Regards

[matthiaskrikey]

Hello, A projected 2D depth map does not encode the texture information. The depth at each pixel only represents the distance to the camera. As you can see in the paper (e.g. Fig.3b), the depth maps do not resemble the typical color input you would get from an RGB camera. Hope this answers your question. Thanks,

[xiaochongs]

thanks for your quick reply. about projected depth image, from your reply I think the pixel value is changed according the distance, it is different with the gray image converted from RGB camera.

In order to fit the input size of VGG Face, a depth map is resized to 224 224 and 3 channels. the depth image is 1 channel. how to resize to 3 channels? make the second and third channel to 0?

Thanks Best Regards

[matthiaskrikey]

You can copy the first channel into the 2nd and 3rd channel. That will make a gray value.

[xiaochongs]

hi，

I use a realsense camera to test it. but I see realsense can give a depth image directly. for example: const uint16_t depth_image = (const uint16_t )dev->get_frame_data(rs::stream::depth);

so can I use the depth image directly? I am confused again what's the difference about the depth_image and the 2D depth image orthogonally projected by facial point cloud.

thanks.