[CentralOH] python blur detection

[Eric Floehr]

Eric,

On Tue, Sep 8, 2015 at 8:57 PM, Eric Miller <miller.eric.t at gmail.com> wrote:

> the "amount of change" problem seems to be the lowest hanging fruit.
> Something like:
>
> - identify an x/y mask that defines the largest rectangular area that is
> 100% sky.  This prevents unwanted changes in light over time on non sky
> objects (trees/buildings/etc) from contributing noise to the amount of
> change calculation.
>
> - using PIL or similar, iterate over every pixel in the mask, avg them
> out, and build a dict of frame #'s to RGB avgs: { '0001' : [10,20,30],
> '0002' : [20,30,40] }.
>
> - compare avg RGB vals for first and last frames to establish start and
> end RGB values. (or skip this and use absolute 0,0,0 = 0%,  255,255,255 =
> 100%)
>
> - compare avg RGB vals for each frame to the one previous, to establish %
> change (relative to total determined in previous step)
>
> wait a minute...didn't you do this already? Like exactly this, lol?
>


Thanks for the great ideas... I am doing similar, in that I'm stacking
images using pillow's lighter() and darker() methods, but not trying to
quantify a change. I know that there is a method to generate an image
difference, specifically ImageChops.difference(), but that results in
another image. I was wanting to somehow quantify that into a number so I
can determine if there is "a lot" or "a little" change between images.

Cheers,
Eric
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/centraloh/attachments/20151019/75aa64b9/attachment.html>