Press "Enter" to skip to content

Ex-Canon boffins seek Land of No Returns

Four Australian imaging scientists, formerly with Canon’s Australian research unit, CISRA, have set up a technology-based start-up, Bandicoot Imaging Sciences. In just six months or so they already have a viable product to offer the world: A new way of viewing cloth and other materials online which gives the viewer an interactive and realistic experience of what it looks like in different lights and from different angles.

Canon quietly closed down Australia’s leading imaging science research facility earlier this year – a loss in itself, but in this case one that has led to an exciting opportunity in the field of technological entrepreneurship for the four former CISRA scientists – Dr David Monaghan (managing director); Dr Matthew Arnison (technology director); Dr Peter Fletcher (chief scientist); and David Karlov(chief engineer).

‘CISRA folded earlier this year and left four of us in a position where we have been able to take some time to establish a start-up company and get into some deep tech,’ David Monaghan told Inside Imaging.

‘The big market opportunity we see is online retail fashion websites,’ he said.

David explained that handling returned goods is a major expense for online retailers, with one in three items returned. Often the issue is that a garment is the wrong size, but buyers regularly complain that ‘it doesn’t look like it did on the website’.

The key to Bandicoot’s ‘dynamic material image technology’ is a physically-based model of how light interacts with the material.

How it works:
Three elemental components can be used to model how light reflects from a surface:
Bandicoot is using a ‘Bidirectional Reflectance Distribution Function’ (BRDF) to model material appearance, which combines these components to describe how light is reflected by a real surface.

Dynamic images of materials are made using a BRDF model with the following layers:
– Diffuse colour: The underlying colour of the material;
– Surface normal: Fine-grained changes in the surface shape;
–  Specular strength: How strongly reflective the surface is;
– Surface roughness: Whether the reflection is glossy, diffuse, or a mix;
– Tint: How specular reflections are coloured by the material.

Bandicoot’s unique image processing technology achieves the extreme precision necessary to show off fashion products in a way which until now simply hasn’t been possible – let alone easy and at low cost. It delivers a solution to a ‘real world’ problem which the Bandicoot team are hoping online fashion retailers will identify and embrace.

‘Bandicoot’s dynamic material images represent the next generation of product visualisation,’ states the Bandicoot website.

While the technology is clearly software-intensive, it doesn’t call for massive hardware expenses. ‘With our technology, all the photographer needs is a DSLR and standard lenses and a flash to take series of images.’

Images files are digitally forwarded to Bandicoot, where they are processed and returned in a matter of minutes. Rather than barely adequate swatches, customers are able to view highly textured, photorealistic 3D renders of the subject material. Online shoppers can see how the light plays on material as they tilt their phone or move their mouse, giving an accurate impression of fine detail such as gloss and texture.

Here’s a conventional photo of a leather wallet, followed by Bandicoot’s interactive version (optimised for desktop):

The other market David and his colleagues are looking at is movies, ads and even computer games which use animation and CGI extensively. One of the weaknesses of CGI is rendering clothing material effectively. The Bandicoot technology changes a process which might cost up to $100K+ – creating materials to ‘clothe’ digital characters – into something far less expensive and, according to David, a lot more realistic.



Be First to Comment

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Our Business Partners