[x264-devel] [PATCH] improved scene cut detection

[Loren Merritt]

On Tue, 26 Jun 2007, Thomas Tzeng wrote:

> Does the scene cut detection is used to support variable size GOP, or
> to speed up the encoding process (like mode decision)?

The goal is to maximize quality-per-bitrate, under some constrained 
maximum GOP length (to allow efficient seeking and a little error 
resilience). Some frames waste more bitrate than others when encoded as 
intra, and some frames are even optimally coded as intra. Scenecut 
detection tries to find those frames, so that the codec won't later have 
to force an I-frame when it reaches the max GOP length, which could cost 
many more bits.

In addition to the bit cost, you also want to avoid I-frames in the 
middle of a scene for perceptual quality: At non-transparent bitrates, 
the artifacts in a keyframe are uncorrelated with the artifacts in the 
previous frame, which makes them more visible than artifacts which stay 
the same.

--Loren Merritt
_______________________________________________
x264-devel mailing list
x264-devel at videolan.org
http://mailman.videolan.org/listinfo/x264-devel