can be a great tool to learn about composition
- Images have real backbones ("structure" or "composition")
- Viewers eyes gravitate toward edge detection; as an artisit, you must use composition to lead your viewer through your landscape
- It is fun, although excessive, to reveal composition with scientific algorithms.
This is a fantastic website for lovers of Art & Science, since it comprehensively reveals compositional design concepts with easy-to-understand visuals. If you want to understand art better, or be a more deliberate designer, check these case studies out ... then apply what you learn.
|Russ's Image Analysis Book|
Here is what we'll get:
(1) a skeleton of features within the primary focus, the "Intensity Skeleton"
and (2) a demarcation of the primary "Contrast Interfaces" that lead the viewer's eyes about the image
To do this, we'll apply a series of operations to our color image.
1) First, we'll isolate the intensity levels by transforming the RGB (red, green, blue) image into HSI (hue, saturation, intensity) map; we'll disregard the hue and saturation for this work and focus on the intensity.
2) Next, we'll apply a median filter to remove the high frequency details since we aim to look at the gross composition (a Gaussian blur).
3) Thirdly, we'll transform the grayscale image (256 gray levels) into a binary image (2 levels, black and white) by common thresholding (we choose a critical gray level that turns all lower to black and all higher to white).
4) Finally, we'll fill-in-holes via a morphology filter.
5) Recolor our binarized image with a Euclidean Distance Map. This will re-shade all black regions with a new intensity dependent on the proximity to the white area. This effectively will make a landscape in which the peaks (the skeleton) can be isolated
6) To isolate the backbones, we threshold our distance map and select values that contain only the peaks.
7) To visualize the backbone of this internal structure within the focus area, we overlay the skeleton atop a version of the original.