Monday, 15 February, 2016. 21:13
Today was the first day of the Electronic Imaging conference. I set the alarm for 07:00 again, to give us time to get some breakfast and then allow me to register for the conference before the first talks began at 08:40.
Last night on our way in we’d stopped to look at the cafe in the hotel lobby, and they had a good breakfast selection, including take away cups of yoghurt, fresh fruit, and muesli, so we went down and bought one of those each. It looked like a lot of yoghurt and not enough muesli, so we also got an extra small tub of muesli to share. This was a significantly cheaper and healthier breakfast than yesterday, and tasted good. For tomorrow we’re doing one better, since they sell small tubs of yoghurt and cartons of milk, so we bought a large packet of Grape Nuts cereal from Walgreens to have with those. Previously we thought we couldn’t easily do this as the hotel room has no mini fridge, so we couldn’t keep milk or yoghurt cool. They also have paper bowls and plastic spoons in the cafe, so we’re all set.
After breakfast I helped M. sort out the trains she needed to catch to go visit Berkeley during the day, then I went down to attend the conference. Registration was quick, with no queues, and I ran into Nicolas (who used to work with me but left to go to Apple in Cupertino) there.
The first batch of talks I attended was in the Image Quality and System Performance conference. There was an interesting talk on developing a single perceptually based measure of image quality which encompassed a range of different image artefacts, such as blurriness, poor exposure, and so on. The presenter said that to calibrate the scale between different artefacts they had volunteer observers judge, for example, how blurry a sample image had to be made until it was equal in “image quality” to a second poorly exposed image. She said at first the observers complained that they couldn’t make a blurry image the same quality as a badly exposed image. So she showed them a nice sharp image and asked if that was better or worse than the badly exposed image. The volunteers said it was better. Then she used a blur slider to make the sharp image really really blurry and asked which was better, the badly exposed image, or the blurry image where you couldn’t see anything. They said the badly exposed image was better. Then she said, “Well, at this sharp end the first image is better than the badly exposed image, but when you blur it out completely then it’s worse. So somewhere in the middle it must be the same!” And after this pep talk the observers could do the task. I’m not entirely convinced that two different image quality axes can be forced to map on to a single axis like this in a consistent way – and indeed she had to throw out three of her observers as their results didn’t agree with the others.
After a brief coffee break, the late morning session began, which was the first of Digital Photography and Mobile Imaging conference, for which I am on the organising committee. This was also a joint session with the IQSP conference, and the talks were all about measuring various aspects of image quality: image stabilisation performance, image flare, MTF of ultra wide angle lenses, chromatic flare (or “purple fringing“), and using standardised illuminants to characterise sensors.
The historic John’s Grill.
Following these talks was the lunch for the DPMI conference committee, so I was invited. It was held in John’s Grill, a famous San Francisco restaurant which opened in 1908 and was apparently featured as a significant location in the novel The Maltese Falcon (though I’m not sure if it made it into the film). The food was of the steaks and seafood variety, and I chose a broiled salmon, which came with mashed potatoes, broccolini, yellow squash, and hollandaise sauce on the side. Over lunch, the attending committee members, a good 18 or so of us, discussed details for organising next year’s conference, which Jackson Roland from Imatest is going to chair.