Until image recognition improves, the only practical way of searching images, is by searching keywords assigned by humans looking at the images (and maybe computers looking at other metadata).

One of the best ways of making your images work for you, is by assigning complete and accurate keywords to your images, but unfortunately selecting and assigning keywords to images is both difficult and exceedingly boring (to most people, anyway).

I’ve found that a well defined "keywording process" goes a long way to speed up this task, and improve the quality of the keywords I end up assigning to images. As part of my workflow for ingesting images I will:

  1. Copy+rename images to their final distinations, using the techniques I describe here
  2. Go through the images and delete obvious rejects
  3. Assign general IPTC stationary to the images
  4. Assign specific IPTC stationary, if the import includes multiple shoots
  5. Batch assign generic keywords to all images
  6. Assign specific keywords to individual images
  7. ……

I found that selecting the "right" keywords could actually be a difficult task and certainly the result wasn’t very consistent, so I’ve identified a number of fundamental questions, which I try to answer for each image:

  • What does the image “say” to you?
  • What is the image about?
  • What is in the foreground?
  • What is the biggest element?
  • Which element has the most impact?
  • What is the major element regardless of how much of the overall image it physically takes up?
  • If there are several identical objects, include both the singular and plural form of that object
  • Start with the most generic form/word describing an object, and "drill down" using more and more specific keywords

Using a standard list of questions, ensures consistency in the end result.

To Structure Or Not To Structure

When selecting keywords for different images, you’ll soon be faced with two dilemmas:

  1. Should similar images have the same keywords, and if so,
  2. How is the same chain of keywords consistently assigned to images where the same specific object occurs

The answer to the first dilemma is a matter of personal preference, but if you plan on selling your images as stock photographs, I strongly suggest a systematic and structured approach to selecting keywords. Many micro stock agencies will routinely discard images with inaccurate or incomplete keywords.

I ilke the consistency in assigning the same keywords to similar images, hence I invariably end up having to find a solution to the second problem.

The only solution that works for me, is to use a tool that will do most of the work for me. Both of my two current preferred image organizers, Aperture and Photo Mechanic, allow you to search in hierarchies of keywords and when a specific keyword is assigned to an image, these organizers can automatically assign the more generic terms in the hierarchy. Combined that with an extensive keyword hierarchy (e.g. this) this ensures a very exhaustive keywording of images.

Other Places Of Inspiration

The same principles that are applied in selecting keywords, when doing search engine optimization (SEO), can often be applied when selecting keywords to describe an image.

One of the best sources for information about adding keywords to images is Controlled Vocabulary, and specifically the guide on adding captions and keyword Caption and Keywording Guidelines.

blog comments powered by Disqus


26 May 2007