quantify image quality

[jacobrosenthal]

Thinking about trying to quantify image quality in memory io shots
Like blur/focus/light
and multiple faces
'interesting' backgrounds, ie not single color, these represent a possible memory of a location

After some searching a term I found is "nr iqa" (no reference image quality assessment) or blind quality assessment Though most of that research is around finding suitable confession algorithms and % that don't distort too much.

thinking you give each metric a score maybe?
more interesting can float to the top then

frankly though nueral networks are the thing for this nowadays

[jacobrosenthal]

@tradica published some early thoughts
http://tradica.com/memory.io/

"When images (spacial domain) are transformed into component frequencies, we can know what type of blur is dominant.
Answers questions like, did the subject move, did the laptop move, did the phot happen while the laptop was opening ...
three general types, linear, radial and zoom
Nothing's exact and it's likely a combination
the reduction process is a virtual committee
and a stack of binary tests conclude in the selection/results
five to ten tests should determine good images
Quantifying result images of transforms and certain pixel zones of those images can reliably determine if it's blurry
Blurry images can be useful/ Knowing the lens profile can isolate foreground subject from background. If a similar enough image is captured in same environment, the blur data can be applied to the clear image
since there's a bit of motion in the blur image, distances can be modeled.
It can be a bit fuzzy, but once a few tricks are piped together there's some awesome power
just added two links to bottom of that page""

[jacobrosenthal]

Heres some stuff I found around nudity

https://be82ac79-a-aff4c786-s-sites.googlegroups.com/a/dcs.upd.edu.ph/csp-proceedings/Home/pcsc-2005/AI4.pdf?attachauth=ANoY7crVHZz4Sr1Q3dXVtYCcrEWjO5d4tvgvzJx-OKgcX8CSlKzjds4m3e8LtjnPn3XDzCsrdjRuWi6CZLI4xNQwrZ7nH-fw55bxCA1f65Tn5vd6bl66pWkiRCK2FlTx0M1hQG_1V28YjihSX0FjffyzJZa1OhhmS0_RTI4c1dsjs3s2mRLZ_nZYQ9MILc2qIQzDL49JTYUVFOKGECkU72cYKdpIoXkkLF1ycHi4tqfG3fEfTKqAVQc%3D&attredirects=1
https://github.com/pa7/nude.js