Its just my internal guideline, and pretty sure that there will be general disagreement.
A recompression for me is a transcode from one codec to another — even within the same family, where a color-encoding, bit rate, or intra-framing recalculation occurs. To a certain extent, not even the render from a grade application, unless you are changing codec, really qualifies as a full recompression, more like a fractional “generation”. All you are really doing is re-assigning the HSL or RGB values of single pixels, and this doesn’t really change its relationship to its surrounding value fields (slots in the pixel array) at least in a lossless codec. Things do change, however, when you start merging and blending external picture elements — blurs, sharpening, mattes, noise, grain… and again if you start geometric manipulation. That’s a different set of rules.
I don’t necessarily count an export from an edited timeline — for example, all-ProRes-422 clips on a ProRes-422 timeline to create a single ProRes-422 Quicktime to be a full recompression. For me, it is fundamentally a simple re-ordering of the frames. However, there are codecs that exhibit losses quickly because they recalculate values based on long-GOP i-b-p economics, or chroma subsampling. Non-square resolutions also raise more than the usual amount of suspicion as to whether you are getting back exactly what you put in.
When you simply copy a digital file from one directory to another… that’s obviously not a compression generation. If you were to change a few of the pixel values inside a frame… but you don’t change the codec, that’s not really a recompression, either. If you pull a subclip or merge clips, unless the codec depends on the field sequence order for its bit reduction scheme, that’s all thats going to change, not the content or values of the intra frames themselves.
Thats my model, anyway.
Love to hear other opinions.
jPo
“I always pass on free advice — its never of any use to me” Oscar Wilde.