Depth of Field Help

Depth of Field is by definition a scatter phenomenon. Theoretically, every pixel should spread its color into the surroundings. Pixel shaders however work in a gather system, so every Pixel has to take the color from the surroundings. Now the problem here is, you can only know if a pixel can spread its color onto the current one (based on blur radius), if you already sampled it.

So there's 2 approaches: if a pixel has a certain blur radius, just read all samples in this radius.
This means you could miss pixels farther away that have a bigger blur radius. And discard a number of pixels that cannot blur onto the current one (like in focus ones onto blurred area). But it's fast.

Then there's scatter as gather, search the maximum blur radius and discard everything that can't possibly scatter its colors onto the current pixel. This means you have to compute a big blur for the entire image and discard 99 percent of the read pixels. Accurate, but slow AF. It's what you're looking for but it'll cut your fps in half.

All DoF shaders for ReShade work with the first method. Magic DoF and Matso are multipass, where the information if a pixel can scatter color is partially lost so the blur is much more even, at the expense of bleeding artifacts around focus objects because pixels in blurred area lose the information if the in focus one should blur into the DoF area or not.

TLDR: What you want is possible, but slow as hell. All DoF shaders use some sort of tradeoff, for performance. My DoF focuses on minimizing bleeding artifacts, others just blur everything.

PS: If you lower the Bokeh Curve of mine, you get a more even blur.

Well I could supply a scatter as gather DoF. Insane performance hog but if that's what you want...