Posted on May 12, 2011 at 13:47 by Kevin Chung in Technology
Video is an integral part of life. From the moment it was possible to record a sequence of images we have been trying to improve in this area at a high pace. Cameras are getting better at capturing fine detail and recording at high frame rates. Consumers can afford better equipment to watch their videos as technology improves and prices drop. I will discuss good quality video as in the perceived visual quality and the essence of producing, mastering and encoding video.
I see aesthetically beautiful videos all the time. That means rich of fine detail and in a way that it feels organic rather than reproduced. For example when you have the idea the flowers in the video really have an accurate but also organic color. And in a sense you can see the structure of leafs.
But I also quite often see videos that are truly awful to watch. Think of completely deformed video because of severe blocking and other commonly known artifacts. This is mostly due to lack of sufficient bit rate and poor encoding.
“Is there any good reason to produce poor quality video?”
Nowadays, there are cameras -expensive ones- capable of capturing the real world in all its glory. However, most end users never get to see that quality. Studios have the masters, but somewhere at the end of the chain all we get to see of that glory is a derivative of much poorer quality. And that is just plain awful.
Some videos, e.g. music videos, are only released to YouTube, with an occasional broadcast on TV. The quality varies from bad to reasonable at best. Yet, there might still be plenty of people willing to pay to see the high-quality version and enjoy all the original glory.
Of course, I am talking from the perspective of the end user, looking for quality and a ‘rich’ user experience. There can be various other reasons why this desired quality standard is not reached. The issue might be financial or political, forcing engineers to abide to and deliver less quality.
Image enhancement filters
Then there are image enhancement filters, which can do real-time processing to compensate for flaws like blocking, ringing, etc. In my opinion, such filters only exist because a lot of videos are poorly encoded. There is no need for such image enhancement filters at the end of the video playback chain. Think of it. If we stuck to the principles of producing good video you would at best do the filtering during the producing and mastering stage.
This means you should always process video in such a way that it minimizes the loss of quality. From a broader perspective you should take into account format compatibility and how end users are going to watch these videos. And of course you should keep it artistically correct.
The encoding stage
The encoding stage is where it often goes wrong. The use of lossy encoders for video delivery is mandatory. Compression is an art in itself and can yield surprisingly good transparent results for file sizes some 10 to 100 times smaller than the original, depending on the source. The trick is to strike the right balance between quality and bitrate.
In other words you want to use the smallest bitrate possible while keeping that transparency. Encoding entails a lot more than choosing a predefined setting. The encoder itself needs to be designed very well to achieve decent compression and detail preservation.
I often feel the encoding stage gets the least attention while it’s just as important as the producing stage. I admit that to many people, the encoding stage can be quite boring in contrast to producing where you concentrate on video content, video editing and filtering. There is enough expertise in that area, but not in encoding.
Compressionists will all agree your video needs to be encoded properly like I described above. This will make or break your product. What is is the use of spending tons of money on expensive cameras and sets for a video if it ends up looking like -excuse my French- crap?
True professionals will not have good times with Apple’s H.264 encode compared to the world class X264 encoder, simply because Apple’s H.264 encoder supports only a very limited set of features defined in the H.264/AVC standard. It also does not allow much tuning at all.
X264 on the other hand allows a lot tuning, which is a great advantage because each video is different. It is maintained at a regular base and its core algorithm regarding analysis is top notch. Apple’s H.264 encoder needs a lot more bitrate to achieve the same transparency as X264 does.
Assuming video is kept intact at the producing and mastering stage -meaning lossless from the camera and free of artifacts-, encoding is all that can make the difference between poor or good quality video. I would rather see all the post processing features in TV’s and ASIC chips to go away. We don’t need them if everyone did what we are supposed to do.
Classic example
In earlier posts, I have written about my earlier views on interlacing. If everything was kept progressive we would not require deinterlacers. If there is no blocking due to poor encoding or compression, there is also no need for deblocking.
As you can see, if none of these flaws were introduced during the treatment of the video stream, it would only require decoding. That would make the ASIC chip design simpler, bring down the costs and use less power, thereby increasing efficiency.
Post processing
Another thing I do not like is that all of these post-processing features are in some way enforced. Take TV’s for example. I bet if you have never looked into the settings, post-processing features like deblocking, skin color correction etc. are turned on. Assuming that the video fed to your TV is already in a good state, the video will be “over-enhanced” which is also bad.
Whether this happens or not depends on the intelligence of the algorithm. Still results are bound to be suboptimal. The worst of them all is, that this is marketed as necessary for your videos to shine. In my personal view this might help for some videos, but still it is biased anyway.
The example above again illustrates the importance of maintaining the essence and good principles of video. It affects so many things. The way we work with video now is poor when it comes to quality preservation.
I’ll discuss this further in a new post and present a new workflow which should address and solve these problems. So don’t touch those dials, folks. I’ll be back soon!
Kevin Chung is encoding engineer at ODmedia.
