News from Innobits - More information about DVNC

From newsbits issue #1, may 2008

More information about DVNC

This article was written as a result from the interview with Torbjörn Bergkvist. Clearly, many of our customers avoid the DVNC noise cancelling filter, due to lack of information on how well it works. We hope that this can be changed.

In the past we have only heard positive feedback about DVNC. However, for obvious reasons, these responses have only come from people who have used it.

After viewing some of Torbjörns sample movies, we had to ask him why he hadn't used DVNC on them. He told us that he had had prior experience with other noise reduction tools. They had all been more harmful than helpful to the quality of his footage. This made us think that there are probably others with similar experiences and therefore never bothered with trying the DVNC option. We have also seen many ill-performing noise reduction products over the last 12 years.

This made us think about how we could help more customers discover this very powerful tool and put it to good use. Feedback from several reviewers and customers who use DVNC, tell us that DVNC ALONE is worth at least as much or even several times the price of BitVice. In the current version 1.8 you are still able to get it for free when you buy BitVice. That will probably change in a near future, so if you want it for free, you may find yourself in a hurry.

We know, as well as some of our customers, that DVNC can make wonders to the quality of the final .m2v file. It can typically save you 15 - 30 % of the bitrate, at the same or higher quality level. This is huge, but if this doesn't impress you, let's think of it from another perspective. How would you value the potential of putting three hours of video onto a DVD, looking as good as if you just put two hours on it?

The DVNC is based on, and has evolved from, one of our patented inventions. This invention was first presented at the NAB show in Las Vegas, ten years ago. It was a "show-stopper" at the time.
Unfortunately, due to lack of funding, it has not been commercialized into a separate product yet. However, now there seems to be room in the current plans for releasing stand-alone products based on this technology, for both the Mac OS X and for Windows.

Two effects of DVNC

The DVNC accomplishes its magic in two different ways, at the same time. One is the primary effect, which was the design goal. The other one is equally powerful, but we tend to think of it as a very pleasant side effect.

Let's talk about the side effect first. You may be able to notice it almost by accident.

The surprising side effect
The pictures get cleaned up from almost all kinds of noise, especially in areas where the noise is most easily detected by the human eyes. For example, look at a stationary part of the screen, like the background behind a "talking head". If you know what video noise looks like you will find it even in very high quality video. With a camera on a tripod and some kind of stationary background, DVNC can make the background become a razor sharp and completely calm picture, i.e, stop it from moving like it was composed of a million colored ants at war with each other. You will need to see this with your own eyes to believe it. Very few people have actually been able to experience this. So far they are mainly found among our personal friends and some of the experienced BitVice users.

However, untrained viewer may need help in noticing video noise for the first time. I think this may be because most people
A) have never been in a TV studio or seen a nearly noise free analog TV picture before.
B) have been watching very noisy analog TV broadcasts in their homes for so many years that their brains have found a strategy for ignoring the noise.

It was about ten years ago, while thinking hard about why human brains can be so very powerful at ignoring video noise, that we started to unroll the chain of new ideas, which finally led up to this patent.

The designed primary effect
The design goal was to make the filtered video a lot easier to compress, without loosing sharpness, especially in stationary parts. That is also what DVNC does. Rapidly moving parts of the imagery do not suffer from a little blurring (motion blur), because human eyes are not fast enough to detect it. They are able to detect both rapid motion and fine detail, but "fortunately" not at the same time.

It is a well-known fact that noise is a "killer" to all MPEG-2 encoders. That is because there are no good tools in the MPEG-2 toolbox for identifying video noise as being an undesired aspect of the source video. In addition to this fact, it is very costly to encode typical video noise, in terms of bandwidth/bitrate requirements.

MPEG-2 offers great tools for exploiting both spatial and temporal redundancies in a video sequence. Simply put, encoder designers can use almost any tricks they can think of, in order to find similarities/redundancies, as long as it can be coded into a legal stream/file. The basic goal for a good encoder is to put as little redundant information as possible into the encoded stream, preferably just enough to "catch" the small differences within each picture (spatially) and between pictures, which are adjacent in time (temporally).

That sounds very good, you think?
Just wait until you learn what kind of havoc noise can make to these otherwise great MPEG-2 tools. Basically, video noise is hiding all the similarities that an encoder is so desperately looking for to exploit. Add video noise and, gosh, now pixels start changing everywhere, but for no good reason. Not only that, but all the time, too. Now every frame becomes totally different from each other, maybe not so much for the viewer, but as far as the encoder can see. This can be so, even if all frames originate from exactly the same still picture.

To an encoder, the unwanted noise makes each individual picture appear to be much more complex (spatially) and thus "harder to compress" than the corresponding noise free picture. Noisy pictures contain much more fine detail which need to be represented in the coded pictures, regardless whether the noise is desirable to preserve or not.

In addition, noise makes it so much harder for an encoder to find, otherwise obvious, similarities between frames (temporally), which is often even more harmful to the encoding process.

What settings to use in DVNC

Almost any kind of footage will benefit from using DVNC. With very high quality pictures a setting of 1 or 2 should be enough. Very noisy movies may benefit from much higher settings. Try a short clip, of typical footage, using different level settings to find what gives the best cleaning effect on your type of movies.

Computational complexity
DVNC makes use of some heavy duty signal processing, so it is very computationally complex. Quality takes time and it may often be worth waiting for.

Blur filters
Many MPEG-2 encoders apply a mild softening pre-filter, by default and behind the users back, before the pictures are compressed. BitVice is an exception; it never uses this trick-of-the-trade unless the user explicitly asks for it.

A mild softening pre-filter is a good tool to use, as it makes it "easier" for the encoder to keep a low bitrate. But, this advantage comes at the expense of loss in sharpness, i.e. a resulting blurriness all over the pictures. Our philosophy is that the user should be in control of whether this trick should be used or not. That is often the reason you find BitVice encodes sharper than others.

DVNC works very differently compared to blur filters. It does no softening, so will keep the sharpness and help the encoder at the same time.

Text: Roger Andersson


From newsbits issue #1, may 2008 More information about DVNC This article was written as a result from the interview with Torbjörn Bergkvist. Clearly, many of our customers avoid the DVNC noise cancelling filter, due to lack of information on how well it works. We hope that this can be changed. In the past we have only heard positive feedback about DVNC. However, for obvious reasons, these responses have only come from people who have used it. After viewing some of Torbjörns sample movies, we had to ask him why he hadn't used DVNC on them. He told us that he had had prior experience with other noise reduction tools. They had all been more harmful than helpful to the quality of his footage. This made us think that there are probably others with similar experiences and therefore never bothered with trying the DVNC option. We have also seen many ill-performing noise reduction products over the last 12 years. This made us think about how we could help more customers discover this very powerful tool and put it to good use. Feedback from several reviewers and customers who use DVNC, tell us that DVNC ALONE is worth at least as much or even several times the price of BitVice. In the current version 1.8 you are still able to get it for free when you buy BitVice. That will probably change in a near future, so if you want it for free, you may find yourself in a hurry. We know, as well as some of our customers, that DVNC can make wonders to the quality of the final .m2v file. It can typically save you 15 - 30 % of the bitrate, at the same or higher quality level. This is huge, but if this doesn't impress you, let's think of it from another perspective. How would you value the potential of putting three hours of video onto a DVD, looking as good as if you just put two hours on it? The DVNC is based on, and has evolved from, one of our patented inventions. This invention was first presented at the NAB show in Las Vegas, ten years ago. It was a "show-stopper" at the time. Unfortunately, due to lack of funding, it has not been commercialized into a separate product yet. However, now there seems to be room in the current plans for releasing stand-alone products based on this technology, for both the Mac OS X and for Windows. Two effects of DVNC The DVNC accomplishes its magic in two different ways, at the same time. One is the primary effect, which was the design goal. The other one is equally powerful, but we tend to think of it as a very pleasant side effect. Let's talk about the side effect first. You may be able to notice it almost by accident. The surprising side effect The pictures get cleaned up from almost all kinds of noise, especially in areas where the noise is most easily detected by the human eyes. For example, look at a stationary part of the screen, like the background behind a "talking head". If you know what video noise looks like you will find it even in very high quality video. With a camera on a tripod and some kind of stationary background, DVNC can make the background become a razor sharp and completely calm picture, i.e, stop it from moving like it was composed of a million colored ants at war with each other. You will need to see this with your own eyes to believe it. Very few people have actually been able to experience this. So far they are mainly found among our personal friends and some of the experienced BitVice users. However, untrained viewer may need help in noticing video noise for the first time. I think this may be because most people A) have never been in a TV studio or seen a nearly noise free analog TV picture before. B) have been watching very noisy analog TV broadcasts in their homes for so many years that their brains have found a strategy for ignoring the noise. It was about ten years ago, while thinking hard about why human brains can be so very powerful at ignoring video noise, that we started to unroll the chain of new ideas, which finally led up to this patent.	The designed primary effect The design goal was to make the filtered video a lot easier to compress, without loosing sharpness, especially in stationary parts. That is also what DVNC does. Rapidly moving parts of the imagery do not suffer from a little blurring (motion blur), because human eyes are not fast enough to detect it. They are able to detect both rapid motion and fine detail, but "fortunately" not at the same time. It is a well-known fact that noise is a "killer" to all MPEG-2 encoders. That is because there are no good tools in the MPEG-2 toolbox for identifying video noise as being an undesired aspect of the source video. In addition to this fact, it is very costly to encode typical video noise, in terms of bandwidth/bitrate requirements. MPEG-2 offers great tools for exploiting both spatial and temporal redundancies in a video sequence. Simply put, encoder designers can use almost any tricks they can think of, in order to find similarities/redundancies, as long as it can be coded into a legal stream/file. The basic goal for a good encoder is to put as little redundant information as possible into the encoded stream, preferably just enough to "catch" the small differences within each picture (spatially) and between pictures, which are adjacent in time (temporally). That sounds very good, you think? Just wait until you learn what kind of havoc noise can make to these otherwise great MPEG-2 tools. Basically, video noise is hiding all the similarities that an encoder is so desperately looking for to exploit. Add video noise and, gosh, now pixels start changing everywhere, but for no good reason. Not only that, but all the time, too. Now every frame becomes totally different from each other, maybe not so much for the viewer, but as far as the encoder can see. This can be so, even if all frames originate from exactly the same still picture. To an encoder, the unwanted noise makes each individual picture appear to be much more complex (spatially) and thus "harder to compress" than the corresponding noise free picture. Noisy pictures contain much more fine detail which need to be represented in the coded pictures, regardless whether the noise is desirable to preserve or not. In addition, noise makes it so much harder for an encoder to find, otherwise obvious, similarities between frames (temporally), which is often even more harmful to the encoding process. What settings to use in DVNC Almost any kind of footage will benefit from using DVNC. With very high quality pictures a setting of 1 or 2 should be enough. Very noisy movies may benefit from much higher settings. Try a short clip, of typical footage, using different level settings to find what gives the best cleaning effect on your type of movies. Computational complexity DVNC makes use of some heavy duty signal processing, so it is very computationally complex. Quality takes time and it may often be worth waiting for. Blur filters Many MPEG-2 encoders apply a mild softening pre-filter, by default and behind the users back, before the pictures are compressed. BitVice is an exception; it never uses this trick-of-the-trade unless the user explicitly asks for it. A mild softening pre-filter is a good tool to use, as it makes it "easier" for the encoder to keep a low bitrate. But, this advantage comes at the expense of loss in sharpness, i.e. a resulting blurriness all over the pictures. Our philosophy is that the user should be in control of whether this trick should be used or not. That is often the reason you find BitVice encodes sharper than others. DVNC works very differently compared to blur filters. It does no softening, so will keep the sharpness and help the encoder at the same time. Text: Roger Andersson