This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

TDA4VM: LDC parameters

Part Number: TDA4VM


Regarding VPAC LDC, we would like to optimize the similarity between an LDC undistorted output frame VS a CPU openCV undistorted output frame.

We believe we have built a valid mesh table, with the correct data format (S16Q3). Our results are very good using  subsampling factor=0 (tested with region out block size of 8x8, 16x16 and 32x32). By good results I mean the LDC output undistorted frame is very similar to frame that was undistorted using CPU OpenCV.

Question 1: We would like to increase the subsampling factor so we can have better performance on the VPAC LDC, but also by keeping and acceptable similarity. Would you say there is a specific subsampling factor we should reach to be acceptable in a real application usage? I mean by that, is there a gold standard factor for real time usage, or would it be case by case?

Question 2: We noticed there are green/black artifacts along the bottom and right side of the output frame starting subsampling factor=2. We tried different region parameters out block sizes, the artifacts got smaller, but is there something else we should be looking into? Is there any other critical parameters we should consider?

Question 3: What would be an acceptable performance (in ms) to undistort a 4K frame using LDC?

Thank you!