具有高帧间相似度的视频压缩？
我有课堂讲座的视频录制，显然有很多帧间相似性。该视频当前规格为：
720P、H264 AVC、30FPS、2456kb/s，文件大小为920MB。
我可以将 FPS 降低到 24，并且可以稍微降低比特率。但是，这个源复杂性较低的视频是否可以使用其他编解码器压缩得“更好”。意思是，是否有已知的编解码器/“编解码器设置”对于如此低的源复杂性来说表现得非常好。
I have videos recording of class lectures, which obviously has a lot of inter-frame similarity. The current specs of the video is:
720P, H264 AVC, 30FPS, 2456kb/s and the file size is 920MB.
I could decrease the FPS to 24 and may be decrease the bit rate a little. But can this video, with low source complexity be compressed a "lot better" with some other codec. Meaning, is there any codec/"codec settings" known to perform well extremely well for such low source complexity.
如果你对这篇内容有疑问，欢迎到本站社区发帖提问 参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱，如果其他用户或者作者回复了您的评论，将不能在第一时间通知您！
发布评论
评论（2）
当任何编解码器最擅长预测下一张图片的尽可能多的像素时，它都会利用最好的冗余。
实现此类技术的最佳机制之一是多假设运动补偿。你可以阅读这篇论文。 http://iphome.hhi.de/wiegand/assets/pdfs/vcip00.pdf基本上，您可以通过应用大量块的平均值来非常准确地预测任何块。
事实上H.264对此有充分的规定；唯一的问题是大多数实现可能无法充分利用这一点。您可以扩展任何编解码器来解决此问题并发挥最佳效果。您使用的编码器可能不是，但 H.264 仍然是地球上解决您的问题的最佳编解码器。
Any codec would exploit the best of redundancy when it will be best at predicting as many pixels of the next picture as possible.
One of the best mechanism to achieve such techniques is multi hypothesis motion compensation. You can read this paper. http://iphome.hhi.de/wiegand/assets/pdfs/vcip00.pdf Basically you can predict any block very accurately by applying as an average of very large number of blocks.
In fact H.264 has full provision for this; only thing is most implementation might not be capable of fully exploiting this. You might just extend any codec to work on this and make the best. The encoder you are using may not be, but H.264 is still the best codec on the earth for your problem.
我猜这些视频是由相机硬件编码器即时编码的，对吗？使用最先进的 H.264 编码器（例如 x264）可以做得更好。在编码之前去除噪声的一些过滤也应该极大地有助于压缩性，但可能会去除一些细节/类似细节的噪声。
对于基本编码，您可以查看手刹： http://handbrake.fr/
如果您不想深入了解，您可以使用其他 GUI，如 Staxrip 等，或直接使用命令行，带或不带 avisynth 进行过滤。高级选项：http://mewiki.project357.com/wiki/X264_Settings
I guess those videos are encoded on-the-fly by the camera hardware encoder, right? You can do much better with an state-of-the-art H.264 encoder like x264. Some filtering to remove the noise before encoding should also help compressibility greatly, but may remove some details/detail-like noise.
For basic encoding, you may look at handbrake: http://handbrake.fr/
If you wan't to dive deeper, you can use another gui, like Staxrip, etc, or the command-line directly, with or w/o avisynth for filtering. Advanced options: http://mewiki.project357.com/wiki/X264_Settings