Re: Advice on "Matrix Metering" for Cams ?

On 5/4/25 4:16 AM, rbowman wrote:

On Sun, 4 May 2025 01:24:21 -0400, c186282 wrote:
 

SO ... wrote a little app that breaks the images into a 7x5 matrix and
yields 'average brightness'
figures for each segment ... the basic of 'matrix metering'.

 That's sort of like pooling in a convolutional neural network. The image
is broken into 2x2 or similar and either the maximum or the average for
the square is used for the next layer. It's interesting to look at the
images after each iteration. It really sharpens lines and shapes that
would be a smoother gradient in the original image and reduces the
dimensions.
 I don't know if successive iterations would yield something you could use
to set contrast.

   What I need is advice on how to best CALCULATE from
   each segment what the contrast and brightness may
   be and then figure out what to DO about it.
   A real-world image ... sun is bright or dim, there
   are a few really bright PARTS in it and some really
   DARK parts. The "best picture takes such stuff
   into account without getting TOO extreme in any
   direction. Camera-makers seem to have this down
   quite well, but their algos are 'proprietary' and
   even extensive searching has not yielded much
   detail into how they do it (with little micro-
   controllers).
   Gonna have to work on this for awhile ... do lots
   of experiments.