Demystifying video formats

Youtube now supports double frame rate video 50p and 60p so what?

That is actually a legitimate question look at this example here which is a short clip from a trip to Barbados in 2013, this was originally shot on a Sony RX100 Mark II in AVCHD progressive 1080@50p 28 Mbps

If you don’t see the 50p option is because your browser operating system does not support it. You need the latest version of browser and operating system and a machine fast enough plus enough bandwidth. So for Mac this means OS X Yosemite and Safari and for Windows you need 8.1 and IE9.

I hope you enjoyed the clip now check this other one which is a instead shot at 25p with the same camera at 24 Mbps/

I think you can see by yourself which one looks better and it is the 25p clip despite an overall lower bitrate.

There are a number of reasons:

  1. Underwater clips do not have a lot of action as you may think so extra frames go a bit to waste
  2. The encoding which is how the clip is first recorded by the camera is not really that different.
  3. The human eye does a great job at interpolating missing frames anyway
  4. There is not really much more data in the 50p file compared to its 30p rendition
  5. The image quality if you look at a still frame is better in the 25p clip.

There are of course benefits in shooting at double frame rate if you want to slow down the footage 50% speed but for what concerns your normal shooting you would say for that clip you could not tell.

Let’s think about it in simple terms if you have a clip shot at 25p with 24 Mbps you would expect something not quite double but a bit more for 50p instead you only have 28 Mbps. To be more precise you have 22 Mbps vs 26 Mbps video which is 18% more in Sony’s case. So that is not really much information more.

What is more interesting is the structure of the data what follows now is a bit technical but bear with me.

GOPStructure
GOP Structure Row 1 and 3 Sony AVCHD 25p and AVDHD progressive 50p

The first and third rows are representation of Sony 25p and 50p clips. The green bar are I frames that you can think of like a JPEG image, the red bars are P frames or prediction the only contain a delta from the previous frame not a full image.

You can see that in the first row there are 12 P red bar between each green I bar. This means that the GOP or group of picture is composed of a sequence IPPPPPPPPPPPP that repeats indefinitely.

On the 3 row there is a representation of a Sony 50p clip you can see that now there are 23 P frames between two I frames.

So the increment in full frame is limited however if we look at the sizes we see that the I frames in the 25p clip are 12% bigger and also the P frames are smaller.

So in short if you look at the image quality the 25p clip has more information in the full frames as well as for the predicted frames whilst the 50p clips has more frames but overall with less quality.

Which means that unless you are shooting something that is really action packed or you want to do slow motion there is no actual benefit but instead a deterioration when you shoot AVCHD progressive underwater.

Note: if instead we were shooting at higher bitrates for the 50p the story would be different but at similar bitrate it goes as above.

You will also have noticed stream 2 and 4 in the image above I repeat them again here

GOPStructure
GOP Structure Panasonic AVCHD 25p and AVCHD progressive 50p

The second and fourth stream are generated by a Panasonic camera and they look different. You will notice now the existence of frames with the tag B and also that some of the P frames have a green slice.

This means that Panasonic AVCHD implementation has two features that Sony does not have:

1. It has B frames which not only predict future frames from the past frames but can also reference future frames in the prediction (sounds crazy but it works basically the frames are stored in memory before past ones are saved)

2. It has slices for images so on one frame there can be an element of prediction from a previous frame and another element completely newly generated for example if the prediction was completely in a part of picture where there was a lot of movement.

H264 encoding has motion compensation so things that do not change are referenced and new parts are predicted or in this case partially created from scratch.

So the Panasonic encoding algorithm is much superior to the Sony one for AVCHD this explains why a small camera like the Panasonic LX7 could produce video to compete with a larger sensor RX100 with almost double number of megapixels.

What makes me laugh is when photography magazines jump to conclusion on the video quality of a camera shooting a static frame!

Of course if there is not movement the camera with the best IQ in still pictures will prevail however when you record motion all of that becomes somewhat less relevant as compression impacts the quality.

So the more effective compression algorithm of Panasonic beats Sony to the point that even a larger sensor size seems not to matter.

This explains why when you take a real life clip Panasonic cameras perform better in video despite a worst image quality in still images.

The difference between the 28 Mbps and 24 Mbps follows pretty much the same trend of the Sony clips there is not enough bitrate to justify the double frame rate unless there is a lot of action in your clip.

So to conclude if you are shooting AVCHD the normal 24/25p more will have better image quality and will be more suitable to scenes with a lot of dynamic range, will give more colour and contrast. If there is really a lot of action or you want to slow down the clip shoot in 50p bearing in mind that image quality will actually drop if you look at a still frame in isolation.

Underwater contrary to what you may expect things do not actually go that fast and most of the movement is in a specific part of the frame or in a limited part of it so AVCHD 24/25p gives better results.

Finally when looking at a camera for video check for real clips do not look at resolution charts designed for still images as they give very little indications on the quality of your videos. Also if there are any tests make sure those are on the JPEG images that share similar processing engine not on RAW files are you are not shooting RAW video. And finally consider that at similar bitrate some manufacturer have a clear edge on others when it comes to real time compression in our example Panasonic produces similar quality to a Sony camera with overall a better sensor but poorer compression.

Advertisements

2 thoughts on “Demystifying video formats”

  1. hi i am a great fan of your blog as I am consistently amazed by your dedication and attention to detail. Following your blog I have been getting quite good results with my rx100 Mark2 using INON uwl-h100 with dome, especially if it is mounted on orange filter. However I am considering switching to LD or AD mount as suggested by your blog in order to reduce wear on the M67 screw threads. IS there any other option to use orange filter with the UWL-h100 with dome setup eg gel filter over the internal surface of the inon lens or even in the housing itself? Appreciate your advice

    1. If you shoot stills you don’t need a filter when you use RAW for video a distorted dome is not my preference so I haven’t bothered. You can buy a spare part to convert your lens to LD mount without the need to buy a new lens. Ask you inon service dealer

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s