• Pixaria Gallery
• Gallery Features
• Gallery Demo

• AssetDeck
• AssetDeck Features
• AssetDeck Demo

• Further Info
• Tech Specs
• Requirements
• Get Web Hosting
• Sell Photos Online
• About Pixaria
• Privacy Statement
• E-mail Address

• Feeds
• Pixaria News
• Documentation
• Twitter

Image Colour Tagging

Pixaria Gallery contains many useful features and it's often the case that some of them may not be widely known or used. One example is the built in support for searching for images by colour which works by storing a mathematical representation of the predominant colour of an image in the database and matching this against user's selections from the search engine.

When colour searching is enabled in the admin area and images have been tagged with colour information, the search engine makes it simple to find images matching or close to a selected hue which is a great tool for picture researchers.

Although this is a very impressive feature, the major drawback is that images need to be manually tagged with colour information by visiting the image editor screen and picking a nearest match colour from a palette of colours. While this does the job, it's somewhat cumbersome to use and so I've been working on improvements to this feature for inclusion in version 3.0.

Starting with version 3.0, the colour selection palette in the image editor will include five colours that are representative of the predominant hues in the image and which will be displayed below the standard colour palette. As in the screenshot above, this makes it much easier to find relevant colours and tag them in the database.

The colour detection system I've built will be weighted towards brighter colours with high saturation and brightness which should help to pick out features like skin tones, and solid blocks of colours.

At the present time, automatic detection of the most relevant colour for an image is not straightforward using software as there's no way to determine which part of an image is the subject but I'm investigating other possibilities in this area which may yet make their way into version 3.

Comments for this article