Story-ception

When I’m not spending whatever spare time I have glued to a computer finishing assignments, I enjoy taking my camera out an about. For me, photography is a creative outlet and also a means of capturing moments in my life.

As I have refined my skills with the camera, I have also become increasingly aware of how to tell stories through images (as they say, a picture is worth a thousand words). There are many ways to craft a narrative with photography: the subject, the composition (the arrangement and framing of your subjects) and how the image is captured (lighting, colour, contrast etc).

In this blog I will focus on colour tone variation in photos. When crafting narratives through a series of related images, a key consideration for me is the consistency in the colour range, so your audience feels a sense of familiarity as they browse. This is particularly challenging when lighting conditions are constantly changing.

My shooting style is emblematic of high impact colours, though I will process images from the same shoot or event in a similar manner, to achieve that sense of consistency (see the Appendix on colour variation). However not everyone is like me! It would be interesting to do this colour variance analysis on a portfolio where a particular style is chosen throughout.

Using the following examples, I represented the colour space of each image in two dimensions using Principal Components Analysis in order to understand how different the tonal variation in my images can be.

I have had a lot of help in processing my images from this blog:
Color quantization in R

---

These 9 images are all from very distinct events (from top left to bottom right):

1. A two panel image of Cottesloe Beach (WA) by day and at sunset
   
2. My sister’s dog, Genevieve
   
3. Danielle at the recent Coogee Beach Volleyball State Tour tournament
   
4. My bike
   
5. Early starters at Bondi Beach
   
6. Something floral from the Botanical Gardens
   
7. A long exposure of Clovelly Beach
   
8. Bike packing over the Border Ranges with Over Yonder Racing
   
9. Yulia roller racing at the Standard Bowl

As colour space is three dimensional (made up of Red, Green and Blue channels), it’s difficult to visualise in two dimensions. Using Pricinpal Components Analysis however, we can reduce the the number of dimensions from three to two. Using this approach, we can visualise most of the variance of the colour space on the X axis (captured by the first principal component), and the remainder of the variance (the second principal component not captured by X), on the Y axis. Finally, the original RGB values themselves have been mapped to the colour aesthetic so we can identify which pixels have been more strongly represented by which component. These outputs are shown on the ‘Re-visualised’ tab.

Originals

Re-visualised

Insights

The immediate standout is the black and white image at Bondi (#5). In the absence of colour, it acts as a baseline, where the first component is largely explaining variance in black pixels, before a gradual shift towards white pixels represented by the second component.

The shapes themselves are interesting: Some have curves which may suggest a smoother transition between colours (#1, #7), others have well defined edges indicating distinct colour profiles in the image (#6, #8). Some display concentration of colours which may suggest emphasis of particular colours (#3, #4, and #7), while others display a spectrum indicating the presence of a wider colour variance (#2, #9).

Final thoughts

My conclusions from all of this are not your typical ‘insights’. For me, it has been an exercise in describing the colour variance of some of my photos. A little closer the ‘Data Art’ side of the spectrum, than your usual hard, actionable insights. An enjoyable exercise nonetheless.