I am familiar with DC restorer. However, isn't it shifting negative part of the signal to positive level? Best to my knowledge CVBS signals do not have negative part.
Yes,they do!
Unfortunately, from Googling, you wouldn't find that out, which is why I find the Internet a little underwhelming as a resource.
As you have found, analog video signals need to maintain particular set levels for the interconnection of equipment.
All the websites I've found that show video waveforms have them simply shown as a 1 volt peak to peak signal, starting at zero volts at the sync tips, & extending to 1 volt at peak white luma.
It makes it easy to show the relevant levels, but it is equine faeces!
In a standard PAL video signal, blanking is at zero volts, peak white luma is at +0.7volts & sync tips are at -0.3volts.
With NTSC, they distinguish between black level & blanking, with zero level at black, so the positive & negative voltages are slightly different, but not really enough to worry about.
DC restorers (which can shift signal DC components both ways) are normally shown on website descriptions with simple waveforms, but video waveforms are
not simple.
The CR time constant ends up being a compromise between what works best at line rate, & what works best at field rate.
For this reason DC restorer use is mainly in TV receiver use, where its advantage is cheapness.
The device of choice in Broadcast TV is the keyed clamper.
Even with the use of ICs, a keyed clamper is a reasonably complex device.
Why do you have to feed the DC down the coax?
It is possible to buy cables which have audio & coax cables in the same sheath.
They are a bit more unwieldy than normal coax, but the audio conductors are usually big enough use for a DC supply connection,& would allow you to dispense with both the "Bias T" & the need to reset the video DC reference level.