Novel Portraiture: Color Theory & Digital Imaging
Color theory encompasses a multitude of definitions, concepts and design applications — enough to fill several encyclopedias. The three basic categories of color theory include the color wheel, color harmony, and the context of how colors are used. When creating images using Deep Dream Generator (image style transfer) I usually begin with color harmony and context. I also use compositing in Photoshop to combine the results I get from the machine. Image style transfer is a subset of deep learning, or artificial intelligence.
For this portrait, I was inspired by flowery wallpaper, specifically the color you see clearly on the subject’s lips: red + yellow + blue. Some people would simply say orange but I’m looking at a balance between the three colors, which led me to generate two images to layer together (composite). Combining or compositing multiple image together has opened a world of possibilities. In this example, the first image had too much black but I liked the lips, so I made another image using a custom “gilded” style with more of the colors I wanted. This is the result:
There were interesting but weird lines in the background of both images, so I used the colors I liked (ex. in the lips) to create a new color layer for the background. This brings me to context or how one color behaves in relation to other colors. I’ve been exploring specular reflection with darker skin, so I added a gold flower pattern to a deeper shade of orange to compliment the subject in the foreground. Compositing is fun but I don’t always use Adobe Photoshop. Sometimes all it takes is one pass using Deep Dream Generator.
There are so many possibilities. That’s what I enjoy most about the process. Sometimes more human intervention is needed and other times it’s helping the machine generate the best version of the image. It takes a lot of practice to know what to do, no matter what tool is used. The latter portrait was generated using an image style I created and uploaded from Happy Color.