In theory, I know a little MPEG4, but it seems to be forgotten, I have to re-see MPEG4, and some people tell me that H264 is definitely better than MPEG4. The result is sent to see H264. After reading a white paper, I have to replace it. H264 boasted:
1. Intra Prediction: The introduction of tricks from surrounding pixels
2. Interped: cut the macroblock, cut, divided, and claimed to predict better. Although the motion vector increases, it turns into Delta
3. TRANSFORM AND quatization: Deceived, there is a seminor on lecturer tells me that there is Wavelet in H264. . . As a result, it seems that it is not seen, but H264 is not finalized, there may be no one. Now that the organization of data in MacRoblock has changed, a block is divided into 16 (or 4) small block, each small block DC and ACs rearrange the organization, and then in coding, the alignment is high. At the same time, improve DCT changes, combine DCT and QUTIZATION. After approximately, it is greatly increased (a variable matrix in the standard), this is still a bull.
4. Reconstruction Filter: Use Filter to take the Blur of the SMOOTH block. There are 2 thresholds in the white paper I'm looking at it.
5. VLC: There are now 2 cases, VCL and Context-Based Adaptive Binary Arithetic Coding (Cabac), first speak VLC. . . . I can't say it. . Cry, anyway, it will change.
6. Cabac: Who can tell me what Context Model is something. . . It looks like to predict, then doing coding.
7. Frame and Picture Managenent: About the display order of Picture, I don't say it, but I can't say anything, my coding theory is rotten. Reference introduces long Term and Short Term, not very fresh, but this trick did play the compression ratio, this is an unconventional. The tragedy will not appear in the encoding efficiency? Is there any bird Codec really made a REORDERING REFERECE PICTURE LIST? Waiting for it. . . .
I have concluded that it is precise. . . .