Tool/software:
Dear TI experts,
Recently I try to modify NSF4 tuning parameters, however it seems to have no difference between stronger and weaker parameters.
Could you help to check attached file that includes raw/yuv/tuning xm/project filel?
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.
Tool/software:
Dear TI experts,
Recently I try to modify NSF4 tuning parameters, however it seems to have no difference between stronger and weaker parameters.
Could you help to check attached file that includes raw/yuv/tuning xm/project filel?
Hi Taylar,
Do you bypass NSF4 on the H/W?
I see that there is difference in the tuning tool project.
You may also check the app note below for NSF4 tuning section for SDR sensors.
Hi Gang Hua,
Thanks for your advise and NSF4 can work normally now. However, the lowlight noise is still obvious even though I set the threshold to match noise level.
Could you recommend other method to reduce noise level in very dark environment. Is there temporal noise filter?
Hi Taylar,
However, the lowlight noise is still obvious even though I set the threshold to match noise level.
Could you please share the raw image as well?
Could you recommend other method to reduce noise level in very dark environment. Is there temporal noise filter?
Unfortunately, DA4 VPAC does not include temporal noise filter.
It can only be implemented in DSP S/W on TDA4.
Further reducing noise in very dark environment by VPAC or sensor may require some tradeoffs, e.g., increasing sensor exposure time (lowering frame rate), lower color saturation to reduce color noise, lower output image brightness (hide noise in dark) etc.
Hi Taylar,
Thanks for sharing!
Noise level is quite high in the raw image especially in the blue channel (lighting may be yellowish at very dark conditions).
e.g., at the same location, SNR for R and G is around 2.4x ~ 2.7x, but only 1.6x for B.
One more thing you may try for this extreme condition is to use higher NSF4 thresholds for B and lower ones for R and G.
That may keep a little more details in the image.
Hi Gang Hua,
Do you mean the threshold from four rows correspond with R Gr Gb B? So, I need to adjust it manually?
//noise_thr_y
{
{ 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384 }, -> R
{ 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384 }, -> Gr
{ 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384 }, -> Gb
{ 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384, 16384 }, -> B
Hi Taylar,
Yes, those are for the 4 different color pixels.
Tuning tool does not let you to set then separately and you will have to manually edit the final xml files.
e.g., export one xml file for B (higher thresholds) and another for RGG, and then merge the 2 xml files manually.