Noise reduction

[VirusKiller]

There seem to be lots of new devices employing the latest noise reduction techniques and I was wondering whether or not certain types of noise reduction are intimately related to the (separate) processes of deinterlacing and scaling.

If so, is it best to go for a single VP and let it do everything (deinterlacing, noise reduction, scaling)?

[JohnWH]

Ideally, for things like mosquito noise reduction, for a film source it should be applied before scaling but after deinterlacing, for a video mode source it should be applied before deinterlacing, I think similar is true for the other types of noise reduction as well*. This would imply that an integrated solution could give best results as it can shift where the noise reduction is applied, although I think the likes of Algolith seem to do pretty well just sitting in interlaced space, but this of course doesn't mean it can't be better imo.

*It also implies that it would be better to apply these algorithms in 420 space, which would mean the DVD player is ultimatley the best place to apply them!

John.

[VirusKiller]

Thanks for that. It's funny - most of my video processing dilemma revolves around SkyHD!!! The need for a Lumagen or VP50 to get 576i, but aging deinterlacing in the older Lumagens and ringing scaling in the VP50. The need for mosquite and block noise reduction with SD material, but neither Lumagen HDQ nor VP50 has much in the way of noise reduction... It's CII, Radiance, or nothing.

Ironically, the JVC HD1 projector has Gennum 9351 VXP processing with all of the features I need (initially) except EDID control for SkyHD and 24fps genlock from 60i material.

[JohnWH]

Sounds like the radiance is the only thing thats going to press the right buttons for you, I guess its only downside is absence of SDI in. Of course we also don't know how well the radiance/9351 will adapt its noise reduction to suit video vs film mode. I did have a converation with Dale Adams about this a while back over on AVS, I think the concluesion was that before/after deinterlace was of marginal relevance, but I'm not convinced that its of zero relevance.

Cheers,
John.

[Gordon @ Convergent AV]

The current Lumagens do actually have mosquito noise reduction. It can't be turned on or off and is kind of a by product of Lumagens scaling algorithm. They actually usually demonstrate this very artefact removal at shows. There is no BAR though.

As has been said before. There will be an SDI to HDMI adaptor for Radiance. I have no details on costs yet.

Gordon

[JohnWH]

Now you mention it I do remember someone saying something about lumagen messing with the filter coefficieints when scaling to do this.

Will they be doing this or using the the 9351 mosquito noise reduction in the radiance?

Cheers,
John.

[VirusKiller]

Now you mention it, I remember someone saying that the Lumagen NoRing scaling inherently reduced mosquito noise...

John - Radiance will be using the more advanced noise reduction algorithms of the 9450...

[VirusKiller]

I ought to add that, although I'd rather not spend £3000 on an external video processor (though I have no doubt that the Radiance will do everything I need and a lot more besides), there is another reason why I'm holding back.

My audio is all-Meridian and I really need to find out how they intend to deal with audio over HDMI before I commit big bucks to another device in the HDMI chain.

Which is a real shame as I'd love the opportunity to beta test a Radiance unit (I'm a software tester). I think my plan is to see what the HD1 can and can't do and grab an HDQ as an interim solution if necessary. There are bound to be quite a few knocking around with Radiance upgrades going on.

[JohnWH]

Originally Posted by VirusKiller

...
My audio is all-Meridian and I really need to find out how they intend to deal with audio over HDMI before I commit big bucks to another device in the HDMI chain.

...

If they're anything like parasound, they'll say they're bringing out an expansion card for the current processors to handle it for a year or two and then turn around and say they have no intention of supporting it in their current range of processors and you'll need to buy a whole new processor when it come out later in the year :-(

John.

[welwynnick]

I believe de-interlacing should always be the first process to be applied to any interlaced video signal. Whether noise reduction should be applied before, during or after scaling I'm not so sure.

I suspect noise reduction should be done before scaling because noise is pixel-level degradation, and will get spread over several pixels after scaling, and therefore more difficult to distinguish from wanted video signal.

Good question, though. never thought about it until now. I think the key is that noise is tackled assuming it's a random signal at pixel level, so temporal averaging will reduce noise without losing detail. In that respect, it's different to film grain in HD DVDs (pet hate of mine...)

In reply to the discussion, I would of course recommend the Calibre VantageHD, which does have EDID control for SKyHd, yet is equally effective with film and video, SD & HD, 50Hz & 60 Hz, and has marvellous noise reduction capability.

BR, nick

[JohnWH]

Originally Posted by welwynnick

I believe de-interlacing should always be the first process to be applied to any interlaced video signal. Whether noise reduction should be applied before, during or after scaling I'm not so sure.

...

BR, nick

Hi Nick,

the reason for wanting to move NR, specifically MNR and BAR, before deinterlacing on a video source, is that during compression DCT & Quantisation are applied at the feild level (ideally anyway, conversation with Dale implied that this isn't always the case). In addition to this application of things like multi-angle motion adaptive deinterlacing mean that after de-interlacing video material invariably has some "invented" material present making it harder to produce best possible results. The same is true for the conversion of the native 420 video generated by the decoder to 422/444 formats, as during the process of scaling the chroma data you're again messing with it, so ideally for best possible results the NR should be done in the decoder.

The above said I think these are second order effects which can always be at least partially compenstated for post deinterlacing.

Btw, MNR as part of scaling can be mathematically the same as doing it before hand, although also true of post scaling it progressively more difficult to do it right.

John.

[Gordon @ Convergent AV]

What aspects of audio handling do you want to know about in the Radiance? It is modular in construction. They have said that they will update it to hdmi 1.3 spec when the parts are available and that this upgrade will be available for those who have the soon to be released current spec versions. Is there more than that you require?

Radiance has capability to use the gennum noise reduction. It also has lots of fpga and Lumagen could develop their own nr if required. They haven't said they will though and I have no knowledge of whether they intend to. The current lumagen scaliing will be in the Radiance. I have no doubt they will further improve on it though once development gets on the way.

Gordon

Gordon

[VirusKiller]

Gordon, it's the Meridian side of things which is the problem.... Meridian are the anti-jitter kings and I'm concerned that they will say that they need to get hold of the HDMI stream before anyone else has "processed" it, for clock retrieval and anti-jitter purposes. That would mean HDMI switching and a single HDMI feed going into the Meridian "solution" and only a single HDMI output from the Meridian processor. This completely shafts calibration on a per-input basis in *any* VP.

TBH, I'm probably being a little paranoid and I need to give this some more thought. Maybe I'm going about this the wrong way. If Meridian end up producing a solution that is incompatible with 3rd-party VPs, then maybe it will be time to consider other audio processor and speaker options...

Here's a thought though... Is the Radiance capable of passing though an input to the second output unaltered?

[tryingtimes]

AFAIK - all these audio delay devices (felston, CII, Radiance, VP50, etc) pass the audio completely unprocessed, so I think your Meridian will be just fine sitting after the VP in the chain.
I would have thought you'd want it that way anyway - You would want the jitter, etc removed in the device which does the d-to-a conversion.

[VirusKiller]

Originally Posted by tryingtimes

AFAIK - all these audio delay devices (felston, CII, Radiance, VP50, etc) pass the audio completely unprocessed

I thought delaying the audio to sync with the processed video is a standard feature on the newer VPs? My understanding is that extracting clock information and re-timing an HDMI audio signal can be far from trivial (in some cases at least), though I'm not sure about LPCM which is the most important HDMI audio "format" to handle. I'm also unsure about the consequences of clocking when the video part is altered and whether or not video clock information is required to recover the audio "clock".

Edit: See the very informative first post in this thread. Seems as if there are two modes (coherent in which the audio clock does depend on the video clock) and a mode in which the two clocks are separate. Second edit: Clearly this has huge implications for the HDMI chain - if the Radiance can pass the HDMI signal completely unprocessed then this may be what is needed.

http://www.meridianunplugged.com/ubb...=001044#000000

so I think your Meridian will be just fine sitting after the VP in the chain. I would have thought you'd want it that way anyway - You would want the jitter, etc removed in the device which does the d-to-a conversion.

Feeding the VP output into the audio input is clearly the neatest solution in this case. However, if the feed is split to two identical signals at any point - one for audio and one for video, then this is a moot point.


Hopefully I'll get an indication in the next couple of days. I hope John Kotches will have discussed this with Bob Stuart in the Meridian Hitchhikers Q&A session at CES.