deblock -filter 2,2 advisable for low bitrate?

[deets]

hey,
Here is my situation. I am encoding my DVD's to take on a long flight in a few weeks, i have a 1gb card. Im using very low bitrates as the resolution is 368x208 for the PSP, 185. this seems to work well enough, but i recall reading on here somewhere that deblocking using 2,2 instead of my current 0,0 could improve the picture and also increase compressibility.

now my aim is not for DVD quality, its a trade off between file size and quality. the main aim is to get as low a file size as possible without it adversely limiting the enjoyment of the TV. so, a softer picture with less blocks might not be such a bad idea, especially if it it improves compressibility.

I just wanted to know if this was a wise move. I am currently doing my own tests, but with 3 pass and a slow comp, I didnt want to waste 2 days doing this if someone here had experience which would suggest i might be wasting my time, or if different settings would achieve my goal.

thanks :)

[Kostarum Rex Persia]

Yes you are right. For ultra-low bitrates, deblocking should be +2,+2, or even +3,+3. Also, you should turn on trellis quantization(on value always) and multi-hex option.

Here is my options for bitrate 200 kbps:

E:\PROGRAMI\Divx kompresori\MeGUI\x264.exe --pass 2 --bitrate 200 --stats ".stats" --keyint 170 --ref 12 --mixed-refs --no-fast-pskip --bframes 2 --b-pyramid --b-rdo --bime --weightb --filter 3,3 --subme 6 --trellis 2 --analyse p8x8,b8x8,i4x4,p4x4 --direct auto --me umh --progress --no-psnr --output "E:\video.mp4" "E:\video.avs"

Good work, and live long and prosper,deets.

[Romario]

Yes, I apsolutely agree with you, Kostarum. Deets, you can use the settings from Kostarum without fear.

[Oline 61]

The deblocker is really a trade-off between high detail with some blocking, and lower detail with less blocking. The higher your bitrate to pixel ratio, the lower you can go on the deblocker and vice versa. In the end it comes down to personal preference.

[deets]

thanks everyone :)

does it also improve compressibility or did i misread?

[Oline 61]

The in loop deblocker is implemented in the decoder based on info coded in by the encoder, correct? I don't believe it helps compressibility.

[Manao]

For ultra-low bitrates, deblocking should be +2,+2, or even +3,+3No. Deblocking is inherently adaptive, so there should be no need to raise the strength when the bitrate is lower. The only reason to do so would be personnal preferences, which is subjective.Deets, you can use the settings from Kostarum without fear.It's imho an overkill. Reduce the ref number to 4 or less, use trellis = 1, and reduce me from umh to hex.does it also improve compressibility It depends on the strength. With constant quantizer, usually, deblocking +2 gives the smallest size, and deblocking -2 the highest psnr. Deblocking 0 is the best tradeoff ( psnr/size ), but not by much ( the choice between -2 .... +2 is _really_ subjective, anything outside that intervalle should be chosen knowingly, for even more subjective / specific reasons ).

[imcold]

In-loop deblocking filter is implemented in the encoder.
After encoder encodes the frame, it decodes it and uses it as reference for motion estimation. The idea behind the filter is to deblock this reference picture before motion estimation (I think that's why it's 'in-loop') to get rid of artefacts and to have a cleaner, better picture for motion estimation, thus saving some bits and improving quality.
Of course, it has to be implemented in decoder too, because it needs the same deblocked picture as reference for proper decoding.
...uff I hope I made myself clear and that this was helpful ;)

[Sirber]

Yes, I apsolutely agree with you, Kostarum. Deets, you can use the settings from Kostarum without fear.no --8x8dct ? why?

[Inventive Software]

8x8 DCT is high-profile. AFAIK the PSP doesn't like High Profile.

[IgorC]

Instead of remove all details by inloop filter you can probe AQ. AQ is far smarter than simple wash out details by inloop. x264 AQ is good at low bitrates. just need to find values of strenght/sensivity for source.

[deets]

C:\Program Files\x264\x264.exe --pass 3 --bitrate 185 --stats "2pass.log" --level 1.3 --bframes 1 --b-rdo --bime --weightb --filter 2,2 --subme 6 --trellis 1 --analyse p8x8,b8x8,i4x4,p4x4 --me umh --progress --no-psnr --output "" ""

thats my current command line, sadly limited b frames due to a bug in the psp software.

i finished my test encodes today, i think i prefer 2,2 in this situation. The loss of detail is countered by the slightly less blocky picture, which is what would distract me from just watching the darn programme :P

been interesting to read the comments though, thanks everyone :)

[shon3i]

x264 AQ is good at low bitratesI agree with you

[Romario]

x264 AQ is definetily good at low or ultra-low bitrates, you should try to use it.

[deets]

whats AQ?

[Kostarum Rex Persia]

AQ is Adaprive Quantization patch for x264 builds. You can use it if you download ChronoCross builds from http://www.chronocrossdev.com/apps/x264/

Hold on, I suggest to download that build later, because ChronoCross isn't, yet, compiled build 465. Build 465 is the latest x264 build.

[foxyshadis]

They're only important if you need strict bitrate control. Anyway, 438 is apparently the last one with fully working AQ, in recent builds it isn't working. Wait until we investigate.

[deets]

sounds interesting, i shall keep an eye on the situation :)