I shot this in order to work on the CC of elphel footage in low lights conditions
I found that recording streaming elphel footage at triggered mode, do cause some kind of slowdown over time, it doesn't seems to be influenced on what gets recorded, maybe my actual computer is way too slow to record it correctly while streaming RGB at 80% jpeg compression? i still have to test and study a lot more to reproduce the error...
this is awfully dark because this is a low light test, in the worst case scenario for the camera, no gain, closed shutter (90º) only 136 watts in total in order to light the first shot (and the subsequent others... ) and WB badly done (no colour gains --> no blue)
here is the nodes I used for every shot:
- Image: obviously... - Lens Distortion: the amount of lens aberration to correct, the default value for the default Elphel lens is around "-0.045" - Final X resolution: in my case since I’m looking for a 2.35 format in a 720P setup, my resolution is 1280x544 so the value here should be "1280" - Original X resolution: in my case since I’m shooting at 16:9 and doing pixel binning/scaling my original resolution is 1296x736, so the value here should be "1296" - Translation Y axis: in order to compensate for the internal image within the shot
- Image: obviously!!!
within this group of filters: first I correct the luma/chroma with all the resolution still intact, I then proceed to correct the lens aberration (not the chroma separation though ), then there is some translation correction in the Y axis (optional), and then I rescale the whole shot to the delivery resolution maintaining the aspect ratio ( there is another step in the shadow which isn't reflected in the filters here, I setup Blender render resolution to my delivery resolution )
in order to get the proportion for scaling the whole shot to the correct measurement I do the following: 1280/1296 = 0,987654321 this value is used "relatively" in both X and Y axis in the scale filter.
I'm shooting with the camera set up at 100% horitzontal sensor resolution and 75% vertical sensor resolution (or the number that closes matches this in the camera internal algorithm) this results in the most wide 16:9 the sensor/lens can deliver
then I do pixel binning / scaling losing half the resolution my recording resolution is 1296x736 which is really closer to 1280x720,since I'm looking for a delivery format at the 2.35 aspect ratio (this is pure 2.35, not 2.40 or 2.39) which can compress correctly (both X and Y should be multiple of 16)
I've discovered this resolution 1280x544 which is: 16x80 x 16x34 so that's what I'm using in order to get native 2.35 images at 720 resolution
unfortunately the original resolution doesn't have enough resolution in order to use the whole 1280x720 resolution with non-sqaure pixels (so at least I could have more vertical resolution for the same final image )