Revision as of 22:02, 4 November 2013

Compression & Codecs in video

(File)Format or Container is the shipping box, codec is the goods inside the box.

determines the way audio and video streams (and other metadata) are stored inside the file
Examples are :
- .avi (Microsoft -> OLD)
- .wmf (Microsoft -> Unflexible)
- .mov (Apple -> supports a lot of codecs)
- .ogg (Open Source -> but not very well supported by editing software)
- .mkv (Open Source -> but not very well supported by editing software)
- .mp4 (works on virtually all platforms)

intra- vs. interframe compression

Why do we compress?

jpeg compression

Types of compression

Intraframe compression
- Compresses each frame (so similar to image compression, with for instance JPG)
- No effective size compression
- High Quality
- Easily readable by edit software
- Apple ProRes, DNxHD (Avid)

Interframe Compression Looking at delta Frames
- Compress BETWEEN frames
- effective size compression
- Not easy for edit software
- Use for sharing -> final output
- MPEG-2 (DVD's, hardware players), DivX (downloads), H264 (currently has best quality vs. file ratio)

Run length encoding > Looks for identical segments (often lossless)
Transform coding > Most used (jpeg, DCT, wavelet) (selectively removes ‘non-relevant’ detail from the image).
Chroma subsampling >Selective loss of resolution (4:4:4 vs 4:2:0)
Reducing bit-depth (reducing colour accuracy)
Fractal compression (replacing pixels with mathematic approximations)

FYI: Uncompressed HD video = 400GB per hour More Info [1]

There's no solution to the question 'What is the best setting for my video?". All depends on the platform you will deliver to!

Download Finished

@@ Line 15: / Line 15: @@
 [[file:interframe.gif|600px|right|thumb|intra- vs. interframe compression]]
-==  Codec ==
+==  Encoding/Compressing ==
 Why do we compress?
 * Image- and videofile sizes reduced
@@ Line 39: / Line 38: @@
 ** MPEG-2 (DVD's, hardware players), DivX (downloads), H264 (currently has best quality vs. file ratio)
 <br>
+=== encoding techniques ===
+* Run length encoding > Looks for identical segments (often lossless)
+* Transform coding > Most used (jpeg, DCT, wavelet) (selectively removes ‘non-relevant’ detail from the image).
+* [http://www.knotless.com/1/post/2012/07/the-mysteries-of-chroma-sub-sampling.html| Chroma subsampling] >Selective loss of resolution (4:4:4 vs 4:2:0)
+* Reducing bit-depth (reducing colour accuracy)
+* Fractal compression (replacing pixels with mathematic approximations)
 FYI: Uncompressed HD  video = 400GB per hour
 [http://www.axis.com/products/video/about_networkvideo/compression.htm| More Info]