A real-world problem

The human visual system is incredibly good at locating and analysing written words in the world around us - which makes it easy for us to forget what a difficult task this is.

Powered by Foveola

SceneReader's success in tackling this tough real-world problem lies in its use of the Foveola shape recognition technology, a wholly new approach inspired by Nobel-prizewinning research into the primate visual system.

Foveola provides significant advantages over traditional character recognition methods, allowing SceneReader to operate without time-consuming and brittle pre-training procedures.

SceneReader: Extracting text from images

SceneReader is a new software technology designed to locate, analyse and report alphanumeric text in a broad variety of photographic images, including highly complex images such as street scenes.

SceneReader is supplied as an ANSI-C programming library, which can be used by your own applications to add text recognition capabilities like this.

The SceneReader API is simple to use, with only three inputs:

1. The input image to be analysed
2. A flat-file dictionary containing a list of allowed words for recognition (which can contain arbitrarily many entries).
3. A pluggable "font knowledgebase" used for character recognition. A powerful, general-purpose font knowledgebase is pre-supplied as standard.

In its simplest configuration, SceneReader requires no user pre-configuration or training; it can be immediately applied to your image collection using the standard font knowledgebase and general-purpose dictionary.