colored point cloud streaming

[bmegli]

Continuing #13
RNHVE part of HVS#8

Proof-of-concept already implemented in depth-color branch.

[bmegli]

From Intel tuning-depth-cameras-for-best-performance

The optimal depth resolution is:

• Intel RealSense D415: 1280x720
• Intel RealSense D435: 848x480

From librealsense#7820

• Intel Realsense D455: 848x480

[bmegli]

From Intel tuning-depth-cameras-for-best-performance\

The “LEFT IR” camera has the benefit of always 1. Being pixel-perfect aligned, calibrated, and overlapped with the depth map, 2. Perfectly time-synchronized, 3. Requires no additional computational overhead to align color-to-depth, and 4. Gives no additional occlusion artifacts, because it is exactly co-aligned. The main draw-back is that it will normally show the projector pattern if the projector is turned on

So we should expect worse result with RGB sensor aligned to depth (or depth aligned to RGB sensor) compared to texturing depth with IR.

[bmegli]

From #14 FOV, for D435

Aligning depth to color will:

• decrease FOV (to color FOV)
• degrade the quality of depth (due to alignment)

Aligning color to depth will:

• lack information about color for the wider depth FOV
• degrade the quality of color (due to alignment)

librealsensee align example may be used to see what happens during alignment (cropping).

[bmegli]

From librealsense#3042 YUYV is the native format of RGB sensor data.

Ideally we would want to use YUYV format for RGB sensor to avoid unnecessary conversions.

From libreasense alignment code we see that it is not possible to align to depth with YUYV color format.

From HVE#18 VAAPI supports both YUYV (yuyv422) and RGBA8/BGRA8 (but not RGB8/BGR8) as rgb0/bgr0

The best we can do is:

• use YUYV when aligning to color
• use other format (e.g. RGBA8/BGRA8) when aligning to depth

[bmegli]

In theory alignment should be performed for undistorted images.

From librealsense#1430 D400 doesn't use distortion models (all coefficients as 0).

We consider adding coefficient estimation to the RGB calibration to reduce the distortion (by about 1 pixel at extremes), but at the moment projection without coefficients is the most accurate you can do

As far as I know nothing has changed in this area (undistortion code present but coefficents always zeroed).

[bmegli]

Functionally it is finished. Readme needs some update before merging.

[bmegli]

Ready for merge

[bmegli]

Merged.

The last thing to do is to update how-it-works section on wiki.

[bmegli]

finished