Aside from perceptual phenomena being at play, I don't see much in the way of the noise distribution actually being modulated. It really mostly just blends in using conventional masking as far as I can see, with the real trick probably being to colorize the noise so that one component of it "disappears" when blended over similar colors, while being more pronounced over others. This also lives by tweaking the original colors to have clean gradated areas. And of course as always: Probably tons and tons of layers, each finetuned to a specific setting. If you can't get a nice noise using the filters, you might try to create soem images by forcing them into pallette-based color mode and use the various dithering patterns as a basis for creating textured brushes and fills.
By teal, I guess you mean the mid blue on the arms, but there is a similar texture throughout the image. As a base, I don't know if this would work?
Fill a layer with 50% grey (it basically needs content to run the filters)
Make it a Smart Object because it is Smart to do so, and is probably going to need some fine tuning.
Hit D to set colours to black and white.
Filter > Sketch > Halftone (set Pattern type to Line).
Filter > Artistic > Film Grain
Hue/Sat Adjustment Layer
Curves Adjustment layer.
Adjust everything to suit
Rather than use the Hue sat layer, which is a pain with all the different colours, use a layer above the texture layer with its blend mode set to Color. You'll need to adjust the Curves layer though, or omit it altogether