Kumatora wrote:Also, does this only apply to DIVX MPEG-4?
No. Any type of MPEG compression (MPEG1, MPEG2, MPEG4, etc.) uses temporal interframe techniques. This includes all the MPEG4 'flavors' like DivX, XviD, 3ivX, H.264 (AVC), etc. Same goes for similar video compression like WMV7/8/9 (VC-1), Real Media, VP6/7 and all other modern-day high-efficiency video codecs.
Recommended editing codecs like Huffyuv, Lagarith, DV, and MJPEG all use intraframe compression where each video frame is compressed within itself without depending on other frames for data. This results in every frame being a Keyframe and provides frame accuracy when editing.
On the quality side of the equation is getting the best, lowest-generation source for footage. Lossy codecs like MPEG1/2/4, WMV7/8/9, etc. in addition to being interframe codecs, take another hit by permanently throwing away video data to achieve higher compression. If high enough bitrates are used, the amount lost can be minimized to hard-to-perceive levels or enough so it's not objectionable. But given that most available downloadable material (fansubs and raws) must balance filesize versus quality, there is inevitably some quality lost as a compromise. The more rounds of compression, the worse the resultant video will look.
A good analogy of lossy compression is this: think of a video as a nice flat sheet of paper, say the size of a page of a newspaper. It's easy to read and write (edit) on and stays flat on the surface of a table for accuracy. Unfortunately, it's size makes it hard to carry and move around. So what can we do? Well, we can fold it up to make it more compact (compression). But now we have permanently lost the paper's flatness and the creases make it harder to write on it properly, not to mention read some of it. The more you fold, the smaller the area the sheet takes up, but the more creases you get. What if we wanted to have another flat sheet again? Well, we unfold (decompress) it and we can make a photocopy of the original (such as it is, and assuming that this photocopier is a super-makes-perfect-copies-exactly-as-it-scans-them copier). This copy, is nice and flat, but now we have remnants of the folding embedded in it. We can reduce the visibility of some of these compression effects by adjusting the copier settings (filtering) which may affect the visibility of the writing on the page also. Now imagine if we repeat the compression-decompression-copy steps a few more times. Or re-fold that original sheet in a different way (re-compression).
Now lossless compression? It's like rolling the paper up into a tube. It takes up more space than folding, but doesn't leave permanent creases. You can unroll it and it (and done correctly) is as flat as before the rolling.
And this doesn't even factor in the effects of colorspace conversions...
Additional reference: VideoCompressionExplanation.pdf