How good is HD supposed to be?

Christopher Woods christopher at custommade.org.uk
Tue May 3 12:52:07 PDT 2016


Yes, I could have been clearer. This depends on the deintelacing algorithm.

Aside from the very first frame of a 50p video, which can only ever be the 
first two fields (...well, that or black), every frame after that is 
effectively taking two adjacent fields and saying 'make a frame out of 
those', so each field is a part of one frame but it doesn't just go 1+2, 
3+4 etc.

IF you're making a full resolution frame 1, you'd use field 1 and field 2. 
For frame 2, field 2 and field 3. For frame 3, field 3 and field 4 etc.

Otherwise, it's black and field 1 for frame 1, field 1 and 2 for frame 2, etc.

I think of the process of capturing interlaced video as a constantly 
shuffling (up-down) grill which alternately covers even, then odd, lines of 
the CCD (just for conceptualising, not how it actually works). As you're 
freezeframing that point in time for half of your frame, the next field 
will be slightly advanced in time by microseconds so it's not the same as  
'taking a picture' every 25th of a second.

You're actually taking 50 shots per second and immediately discarding half 
the resolution, relying on persistence of vision and the inherent 
properties of the TV to mask this. The two half-resolution images 
interleave neatly and produce a full resolution image, and do so rapidly 
enough that everything works. You get pseudo-50fps as a happy by-product.

http://www.100fps.com has a good explanation and screenshots of various 
scenarios if you're interested in what raw interlaced video looks like and 
the problems you can have working with it.

(I hate interlaced video.)

200, no, 300 fps 4K video for all! It divides nicely with 25 and 29.97 fps 
standards, it's got the temporal and dimensional resolution, what's not to 
like... Except the transmission and storage costs... But H.265 will solve 
all of that ;)

Chris


On 2 May 2016 8:52:25 p.m. "Dave Liquorice" <allsorts at howhill.com> wrote:

> On Sat, 30 Apr 2016 22:49:01 +0100, Christopher Woods wrote:
>
>> The deinterlacing algorithm is doing no resizing - it's interpolating
>> between the frames and then 'printing' that to 50 progressive frames. The
>> resulting image will have slightly lower definition due to the bob
>> artifacts as it's reconstructing the frame from two sequentially
>> interlaced fields, but it hasn't changed resolution.
>
> Sorry, I'm still missing what is actually going on but I think I'm getting
> there.
>
> Is the first reference to "frames" refering to a frame constructed from the
> two fields designed to be shown at 25 fps? Lets call this F1.
>
> At 50 fps we need twice as many frames so an F1 and an F2. F2 doesn't exist
> and is created by interpolation between the current F1 and the next F1.
>
> --
> Cheers
> Dave.
>
>
>
> _______________________________________________
> get_iplayer mailing list
> get_iplayer at lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/get_iplayer





More information about the get_iplayer mailing list