An interesting approach: take linear RAW images and create a black and white image from them, using it to gather detail and brightness information and, for example, determine sharpness. Then create a second image that adopts the color values - in this image, for example, apply noise reduction and similar mechanisms. After that, combine these two images - take the brightness (and thus the contrasts) from the black and white image and merge the color values from the black and white image. As a result, you get an image with good sharpness and good color, but fewer artifacts from the respective optimization steps (since each optimization is only applied to the elements that can handle these optimization steps without creating problems).