The ringing distortion in broadcasting and media industry: why it matters?

In our posts we have discussed different types of distortion that could affect to the quality in video streaming. Among those distortions, ringing is one of them.

General speaking, we can say that it is very similar to blocking. Because of that, it could be very annoying and could impact in a negative way in the quality experience of user (QoE).

The good news is that this distortion is not very common. The reason of that is because ringing is produced when analogic audiovisual content is digitalized. Anyway, there are some technics that can prevent ringing artifact in the video, such as the use of cosine filter, an exponential filter, vendee filters, and also an adaptive median filter.

What is ringing

But let’s start with a quick explanation of what is ringing distortion and how it occurs. The first thing we have to be aware is the fact that ringing could affect to quality deterioration of audiovisual content and could be done by edge enhancement assuredly. The enhancement-caused ringing effect tends to occur in the pre-processing stage, a previous process to video compressing.

Edge enhancement is an intended process. Its essence is to strengthen high frequency components. But it is important to notice that edge enhancement is not the only cause of ringing effect, although it is true that it is the primary. In fact, some sharpen algorithms can cause outer rings. In that sense, video engineers with a lot of experience can easily avoid such cases, thanks to the fact that edge enhancement process is highly controllable.

Regarding compression, DCT-based compression but also other transform-based compression, such as wavelet-based compression could generate ringing effect in video files.

Compression is also an intended process, as well as edge enhancement. Its essence is to weaken high frequency components. When overdone, it will also cause ringing effect.

Two different effects

The question is how it is possible that strengthening and weakening high frequency components will similarly cause ringing effect.

We must notice that compression-caused ringing effect is very different from enhancement-caused ringing effect. Near the same original edge, the position and size of the enhancement-only caused ringing effect is different from compression-only caused ringing effect.

Compression will often cause “outer rings”. It’s a typical feature of overdone quantization. However, not every edge enhancement algorithm can cause “outer rings”.

In other hand, edge enhancement will always generate haloes. The good point here is these haloes are expected results unless too noticeable. If not overdone, the haloes will make picture look vivid and crisp. That’s the primary reason why only overdone haloes will make audience uncomfortable with this effect.

However, compression-caused haloes and outer rings are all unexpected by spectators. Only with high enough quantization precision, the intensities of haloes and outer rings will be low enough to inconspicuous.

Compared with the whole picture, compression-caused ringing effect only occupies a small number of pixels, so that its influence on objective quality (measured with PSNR) is very low. But it will heavily deteriorate subjective quality.

How to solve it

Those methods involve some post-processing steps to reduce ringing effect after decoding. However, in many application areas, simpler decoder is better.

As we explained, ringing is an effect that occurs specially when we digitalize old audiovisual content. Video industry usually does not care about the complexity and computing costs of encoders, because encoding step is done in the production process. Once a master disc is produced, they can manufacture as many retail discs as they want. So, the production process is not very time-critical, and the encoding need not be real-time.

On the other hand, more complexity and computing costs of decoders bring higher prices and lower popularization of end-user playback devices.

Since compression-caused ringing effect is from overdone quantization, increasing quantization precision is a straight solution. With limited or appointed average bitrate, it is impossible and unnecessary to increase quantization precision of all blocks and all pictures.

There are two methods to reduce ringing. Both are easy to implement with an acceptable computing cost. Also, they are standard output bitstreams:

• Macroblock-based Ringing Effect Reduction. It is only applicable on macroblock-level quantizes adjustable video coding standards.
• Picture-based Ringing Effect Reduction, applicable on picture-level quantizes adjustable video coding standards.

No Comments

Post A Comment