Tuesday, April 24, 2012

Something else for the standards bodies to do

If you watched Scott Israel’s presentation on “Debunking the Megapixel Myth” you know that lens performance is defined in terms of MTF. Unfortunately though for us users of machine lenses, MTF data is rarely made available, and when it is published, it’s in the form of charts that require a PhD in optics to understand.

That’s a problem akin to buying a car without knowing how much gas it uses: as an educated buyer you can make an informed guess, but wouldn’t you feel better making comparisons on the basis of actual numbers?

That’s why new cars are sold with “official” fuel consumption numbers. These are determined under a defined set of test conditions that have only limited relevance to real-world driving conditions, but they permit comparisons to be made between different models.

I propose that similar testing standards be introduced for lenses, and that all manufacturers be required to present the results in a standardized format. The numbers wouldn’t necessarily guarantee performance in a particular application – your results may differ – but they would let the users of machine vision decide if it’s really worth spending $1,500 on a lens when there’s a $200 dollar version available.

Come on Standards bodies – rise to the challenge!

No comments: