Properly exposed, satisfying colors, no visible noise, and busy background in an out of focus bokeh blur.
We can observe some aliasing artifacts however, likely due to line-skipping high speed readout from the sensor. It might be more aliasing artifacts than the iPhone slow-motion, and comparable to other Android phones on the market.
I've been fascinated by this explanation on how calibration was done for film distribution with inevitable deviations due to the analog nature of the process.
In this video, +CineTechGeek shows it consists in calibrating essentially the primaries coordinates. I wonder what the response curve is tho: I suppose essentially linear with a rolloff in highlights? Cool stuff, I'll watch more or those videos to continue learning about it 😊
From a quick look it now seem like videos are upscaled using a basic Bilinear interpolation.
This is still poor compared to Bicubic, Lancoz or even better interpolation as provided by native hardware video surfaces on most chipsets but at least it's not completely insulting like it was before.
Note: tested with YouTube 5.5.27 on Nexus 7 (2013) 4.4.2
The update doesn't use a video surface anymore which leads to several issues:
– Poor bilinear scaling on most devices or no scaling at all (nearest neighbor) instead of the hardware scaling which is usually extremely high quality (lanczos or better) even on low end hardware that comes automatically with video surfaces.
– colorspace conversion issues leading to banding, black & highlight crushing or clipping as reported by a kind follower on twitter.
I get that multitasking is nice, but compromising with video quality is never a good idea.
Note: When you take a screenshot of your tablet displaying a video surface, you get a software conversion that is not what's actually shown on display by hardware decoding and scaling.