The Archiva project has not been moved, and if the outbound book is not today inexplicable night, it is almost the same. . . This week is really fast. . . Summary, I will definitely forget something. . . :
One. Re-looked at MPEG4 outline.
1. MPEG4 hopes to build a MultiMedia system, including local and transote, similar to today's file system, since it is a file system, naturally to define a system mode, that is, DMIF Network Interface (DNI). DNI classmates are just a defined interface. The classmates who do things are transmux layers. This classmates are more qualified. As long as they do, how can they do, the tool he usually use is UDP / IP, MEPG2 TS, AAL2 ATM , H223 PSTN, etc.
2. Since MultiMedia is known, then naturally. Video, Text, Audio, Speech, 2D / 3D Image is all available. These students who have the protagonist in previous standards have dropped a grade, which is known to be Object, managing them is a boss called Scene Description. Scene Description is very cattle, and he can take a stream alone as a text file. I heard that it can be convenient for him to shoot users and make people's horses, let them directly modify the properties of those star objects.
3. Audio classmates don't make progress in my document, or a AAC, there is another Twinvq I don't know. Now I heard that the AAC Group has added AAC Plus, with super advanced *** Bond technology (forgotten), and the code rate is reduced. SPEECH is HVXC and CELP, oh, it is something I don't know much, but the most advanced should be WMA, last year. In terms of artificial sound, first the CODEC of TTS, MPEG4 has a multi-study for artificial intelligence natural lazy, and defines an interface (TTSI). Then Structured Audio (such as a famous MIDI), but MPEG4 can certainly can't call MIDI, so it is used as a comprehensive and general-purpose SAOL and SASL than MIDI.
4. The IMGAE and Animation section are also a subset of video, including 2D / 3D Mesh Texture, Face and Body Animation, and the compression of Mesh seems to use the offset vector and the Affine matrix, Texture uses a cute Wavelet (naturally Scalability's good things, but unfortunately for the compression ratio of VIDEO, it is worth studying), people don't understand, seems to define some predetermined movements and the way to synchronize with the sound; . .
5. Video is my favorite. This time, it seriously considers the Robust during the network transmission. It has made a lot of reply and positioning methods. It is very creative to open the motion vector and DCT factor. RVLC is finally MPEG.
6. There is also the famous Fine Grain Scalablity technology, but now everyone should know. . .
two. 2 Paper about Video Segment.
The first article is a mathematics that I don't understand, I see that the key part is very happy to give up. . . Let's take a Title: Robost Segmentation and Tracking of Colored Objects in Video (IEEE Trans. On Ccricuits and Systems for Video Tech. 2004). I seem to understand, but now I don't understand it, I'm watching At an eye, I remembered it, it seems to make the first frame segment with KMCC (input is the color of CIEL space, avoiding the influence of LuMA), then designed the heuristic TRACE method to go to TRACE to take the segment of each frame. More, use Bayes classification), and others are not the same, he didn't combine segment into Object, but TRACE was talented, according to him said that the effect is good. The algorithm of the combination also claims that it is original, and the original often means heuristics. . . Everyone is here :). Sticking Title: Video OjBect Segmentation Using Bayes -Based Temporal Trracking and Trajectory-Based Region Merging (IEEE Trans. On Ccricuits and Systems for Video Tech.2004).
Two paper are actually unrelated with my research, and I have seen it at night. . .
Ok. Other, forget it. . . .
