When I was in JianShui, I took a picture of a red flower in the rain in front of a traditional building:
Even though I liked the composition, the picture felt a bit pedestrian. I tried converting it to black and white, using the channel mixer to get the red flower almost white, which I liked much more immediately:
Then I got the bright idea to open both pictures as layers and go wild with layer modes. The lighten / darken modes (lighten only, screen / darken only, burn) didn't really do anything interesting, but HSL colour (colour picture upper layer) ended up with a much nicer colour picture than the original (I like how you can see the rain drops clearly). This makes sense, HSL colour takes the colour and saturation from the top layer and applies it to the lightness of the bottom layer. I had previously used the channel mixer to lighten up the red from the flower. That is a really interesting option, I will keep in mind for future pictures.
Now for really experimental: Difference mode calculates the RGB values from the top layers minus the ones from the bottom layer. In case there is a negative number in converts it to positive. For example most of the red flower has values around (180 red, 5 green, 0 blue) and (180 red, 180 green, 180 blue) in the black and white image. The outcome image then had a teal flower with (0 red, 175 green and 180 blue). Any similar colour on both pictures (such as the white sky) would turn to black (0,0,0). It's a cool effect:
Now divide has even more complicated calculations it takes the RGB values from the upper layer * 256 and divides it with the RGB value of the lower layer + 1. The result is that similar colours go white and thanks to the channel mixing I did with the black and white picture only the red flowers and green leaves were left. I really like this one: