Friday, August 05, 2016

Image Quality vs Speed in Mobile Phone Cameras

A brave PhD candidate from University of Vaasa, Finland, presents his thesis for public discussion a week before the thesis defense scheduled on Aug. 11, 2016. The thesis "Benchmarking of Mobile Phone Cameras" by Veli-Tapani Peltoketo is available for download and "defines a new benchmarking method for mobile phone cameras. The benchmarking combines different image quality and camera performance metrics to a comparable, single value benchmarking score. According to the research, heavy image quality algorithms used nowadays may slow down functionality of mobile phone cameras. Moreover, the research highlights the main principles to compare mobile phone cameras."

"Nowadays an image captured by an imaging sensor is practically rebuilt again in a post processing pipeline of a digital camera system. The doctoral thesis reveals cameras which image quality features are superior but correspondingly the same camera can be one of the slowest of the comparison."

"Even though the image quality of a camera can be extremely good, slowness of the camera may prevent to capture the right moment and the result is as same as an image with poor quality: the image is not used", says Peltoketo.

"The research introduces also main principles which should be taken into account when benchmarking scores are implemented for mobile phone cameras.

Firstly, all measurement methods and equations have to be public. Used metrics, measurement methods, and especially equations may include intentional or other weights which can bias the result.

Secondly, different measurement environment should be used when mobile phone cameras are benchmarked. Especially low light environment is very challenging to mobile phone cameras. Image quality and speed results may vary significantly in low light environment.

Finally, a perceptual image quality should be taken in to account when benchmarking is done. A color saturation is a good example of differences between perceptual and objective metric. An image looks better, if the color saturation is artificially increased. However, objectively the image includes a color error.

One may ask which is the best mobile phone camera? Peltoketo notes that it is quite difficult to give an unambiguous answer. However, if the results of the thesis give some answers of cameras available year 2014.

“The best image quality was measured from Lumia 1020 and Zopo C2 cameras whereas the speed performance was best in iPhone 5s Samsung S3 cameras.


7 comments:

  1. Surprising this is adequate for a PhD - no way it would have flown to develop a new image quality test and get a PhD when I was in grad school. I guess the world is running out of new ideas :)

    ReplyDelete
  2. I really don't think this is a PhD work. Structure looks more like a some kind of degree thesis rather than PhD one.

    ReplyDelete
  3. This comment has been removed by the author.

    ReplyDelete
  4. I would like to say Yole is better...

    ReplyDelete
  5. Some things are way too complex to reduce them to one single score value. We all tend to think more and more in "Like" Buttons. The Brexit should have been a warning to all of us, especially in such a scientific field...
    -DKF

    ReplyDelete
  6. Best of luck with your viva, Veli-Tapani.
    Putting forward your thesis (or being able to) for public analysis will always attract some naysayers. Mine is still under lockdown from the industrial partner some 6 years after publication!

    ReplyDelete
  7. I decided some comments should not last forever and so I have decided to delete my earlier comment.

    ReplyDelete

All comments are moderated to avoid spam and personal attacks.