Having finally seen the light - literally - I took a departure to the crisp, clear, and bright world of daylight photography.
Okay, I lied; that was actually taken in the dead of night, complete with obscenely long exposure times and a loyal phone to keep my attention whilst the poor camera slaved away. In my previous shoot, I had tried to combine two entirely different levels of lighting, with less than satisfactory results- so, naturally, I decided to get a bit more creative (those who know me will be running for cover right about now).
As before, I took two separate groups of photographs for the foreground and background, setting the exposure so as to retain as much color as possible. The foreground consisted of a mere two layers, combining hue via sigmoidal addition and value by multiplication and scaling - a process which resulted in the rich, saturated colors of the leaves and dirt, while maintaining the contrast of shadows and relative color differences of the scene.
After the less-than-satisfactory diffusion of lighting in my last shoot's result, the background is where the process got a pit more interesting; during multiple longer exposures, I adjusted the focus of my lens, eventually selecting the shots where I started slowly and increased speed towards the end. I imagine this was responsible for the overly-exposed portions along the upper part of the tree, which I happily traded for more diffused lighting and a sense of bloom.
Now, although these two processes resulted in an appealing characterization of their respective scene, the two composite photos contained virtually no information about the remaining portion - that is, certain parts were entirely under or over exposed, yielding a poor photo individually. Having previously struggled with ISO and image noise, I wondered if this could be used to my advantage.
At the first layer's settings, I retook several photos at the highest ISO my camera would go, taking a short enough exposure so as to not capture the darker portions of the scene. Each photo was then processed by an inverted noise-selection kernel (ie, the noisiest parts of the scene are black, and the most coherent, white). The result was then blurred, re-synthesized, and multiplied with the inverted lightness of the foreground, giving something like this:
The result was a reasonable approximation as to the "accuracy" of each pixel, telling me exactly how I could combine my photos to preserve detail and dynamic range. Using the above as a simple blend mask on the foreground and background layers yielded this post's main photo, a result I'm much happier with than my first shoot.
Okay, I lied; that was actually taken in the dead of night, complete with obscenely long exposure times and a loyal phone to keep my attention whilst the poor camera slaved away. In my previous shoot, I had tried to combine two entirely different levels of lighting, with less than satisfactory results- so, naturally, I decided to get a bit more creative (those who know me will be running for cover right about now).
As before, I took two separate groups of photographs for the foreground and background, setting the exposure so as to retain as much color as possible. The foreground consisted of a mere two layers, combining hue via sigmoidal addition and value by multiplication and scaling - a process which resulted in the rich, saturated colors of the leaves and dirt, while maintaining the contrast of shadows and relative color differences of the scene.
After the less-than-satisfactory diffusion of lighting in my last shoot's result, the background is where the process got a pit more interesting; during multiple longer exposures, I adjusted the focus of my lens, eventually selecting the shots where I started slowly and increased speed towards the end. I imagine this was responsible for the overly-exposed portions along the upper part of the tree, which I happily traded for more diffused lighting and a sense of bloom.
Now, although these two processes resulted in an appealing characterization of their respective scene, the two composite photos contained virtually no information about the remaining portion - that is, certain parts were entirely under or over exposed, yielding a poor photo individually. Having previously struggled with ISO and image noise, I wondered if this could be used to my advantage.
At the first layer's settings, I retook several photos at the highest ISO my camera would go, taking a short enough exposure so as to not capture the darker portions of the scene. Each photo was then processed by an inverted noise-selection kernel (ie, the noisiest parts of the scene are black, and the most coherent, white). The result was then blurred, re-synthesized, and multiplied with the inverted lightness of the foreground, giving something like this:
The result was a reasonable approximation as to the "accuracy" of each pixel, telling me exactly how I could combine my photos to preserve detail and dynamic range. Using the above as a simple blend mask on the foreground and background layers yielded this post's main photo, a result I'm much happier with than my first shoot.
Comments
Post a Comment