[FFmpeg-user] Programmatically detecting 'busiest' parts of a video

[Rob Hallam]

Hi Mark, thanks for your suggestion too.

On Tue, 6 Aug 2024 at 23:34, Mark Filipak <markfilipak.imdb at gmail.com> wrote:
>
> Hey Rob,
>
> On 06/08/2024 08.40, Rob Hallam wrote:
> > Hi,
> >
> > I'd like to programmatically detect the 'busiest' parts of a video- ie
> > the most visually active areas. I am leaving audio aside for the
> > purposes of considering this.
>
> If by 'busiest' you mean greatest frame-to-frame image movement and if by 'parts' you mean temporal
> frame sequences, then you might want to extract motion vector lengths and 'scoreboard' them over
> time to find the frame sequences that have the longest MVs. But if by 'parts' you mean the x-y
> picture areas with the most movement, MVs could be mapped. Good hunting.

Yes, you are correct with the first one- the idea is to find the times
(time ranges) for which there is the most visual 'activity'.

I actually don't have a good concrete definition of activity, other
than a time range which is unchanging or has low change is low
activity; and one which has rapid changes is 'high' activity. I don't
need much in the way of precision.

Thanks for the tip about motion vectors, I'll look into how I might go
about extracting those and see what I can find.

If others have insight too, please feel free- someone must have done
something similar, but I'm not sure of the technical term for "video
activity" to check literature!

Cheers,
Rob