Revision history [back]

disparity_view expects disparities to have type float, so you'll need to convert your uint8_t disparity values. In your DisparityImage message, resize image.data to width * size * sizeof(float), and get a float pointer to your data buffer: float* ptr = reinterpret_cast<float*>(&image.data[0]). Then copy the disparities over one by one. You may need to scale the uint8_t disparities down by some factor if the custom SDK does sub-pixel correspondence like OpenCV's block matcher. The float disparities are in pixels. Don't forget to set image.encoding = sensor_msgs::image_encodings::TYPE_32FC1.