I am trying to simulate a monochrome camera, but have noticed in my tests that the L8 and L_INT8<format> options result in a grey image that only considers the RED color channel of the rendered scene. This could lead to pretty misleading results, as shown in the image below, where the purely green and blue objects are interpreted as black (0), and the pure red background is interpreted as white (255):
bpp
Type: SystemUInt32
The bit depth (bits per pixel) of the raw bitmap.
I have a hunch that the renderer is producing an RGB image by default, and the L8 option is just taking the first 8 bits of each pixel, i.e. the Red channel only.
Certainly more efficient than averaging over the 3 channels for each frame, but it feels a bit misleading...
I am trying to simulate a monochrome camera, but have noticed in my tests that the
L8andL_INT8<format>options result in a grey image that only considers theREDcolor channel of the rendered scene. This could lead to pretty misleading results, as shown in the image below, where the purely green and blue objects are interpreted as black (0), and the pure red background is interpreted as white (255):Looking at the source code, it looks like L8 and L_INT8 set
bppto 8, which is used byFreeImage_ConvertFromRawBits()as the bit-depth:I have a hunch that the renderer is producing an RGB image by default, and the
L8option is just taking the first 8 bits of each pixel, i.e. the Red channel only.Certainly more efficient than averaging over the 3 channels for each frame, but it feels a bit misleading...
Related Gazebo Answers question here: https://answers.gazebosim.org/question/27793/how-to-simulate-monochrome-greyscale-camera-in-gazebo/
System Info: Gazebo 11.9.0 binary install Ubuntu 20.04