I am using the Structure Sensor with OpenNI on Windows. I have built an application which outputs the IR data from the Structure Sensor as a cv::Mat via OpenCV. However, the outputted IR image is very dark (almost completely black unless pointing it directly at a light in which case you see the dim outlines of the light). By scaling each pixel’s grayscale value by say 25x, it begins to resemble what you’d expect a proper IR image to look like (i.e. clearly defined objects, etc.).
I have tried my same application with the Kinect v1 and it works as expected without any scaling performed. For the record, I have tried the Structure IR sensor with both PIXEL_FORMAT_RGB888 and PIXEL_FORMAT_GRAY16, both output the same.
Kinect V1 Without Scaling:
What is going on here?