 Image Integration Newsletter |
Putting Quality in the Digital Image January 2008 |
In this issue:
Image Integration at Electronic Imaging 2008 - Paper presentation; demo session
Subjective Testing - If you haven’t measured the image quality of your product, your customer WILL!
Newsletter Archive - Back issues archived
I will be presenting a paper co authored with Dirk Hertel of Sensata Technologies summarizing our preliminary study of camera phone photospace and image quality. As previewed in last month’s newsletter the results indicate a significant lack of correlation between the conditions under which most camera phone images are made and the conditions required to produce high quality images. The paper also explores some of the fallacies of the ‘mega pixel myth’. The paper entitled: Megapixel mythology and photospace: estimating photospace for camera phones from large image sets is scheduled for presentation as part of Conference 6808.
We will also be demonstrating Image Integration’s Image Phi, a tool for estimating photospace distributions of digital images and a pre-release version of a software tool for estimating image quality.
All a customer experiences of your product is its quality, price and ease of use. The quality of the viewed image is of fundamental importance, yet most technical people are extremely reluctant to consider that this perceptual quality is subject to measurement and quantification. Based on 20 years of experience with measuring image quality we can offer support to assist you in designing, performing and analyzing the subjective measurement of image quality. You can contact us at : bror.hultgren@i-2-q.com
The primary impediment to understanding the concept of the measurability of a
subjective response is the belief that since different observers may disagree on
the quality of an image no quantity can be reliably determined. As a
consequence, there is a tendency to fall back on objective measurements of some
physical attribute of the imaging system. This is often misleading since it
ignores the fundamentally non linear manner in which objective attributes
combine to create a subjective response. The other response is to rely on a
singular observer — often the most senior staff member – to render his judgment,
which is subject to his own bias.Most of us are trained in the physical sciences and as such have been trained to utilize statistics as a means to quantify the imprecise measurement of what is believed to be a unique physical quantity. Subjective attributes such as sharpness, graininess and image quality are inherently imprecise. Statistical measurements are fundamental to quantifying these attributes. They really do work AND are simple. The central limit theorem RULES!
While an individual's responses are widely variable the MEAN of a population is
remarkable stable. Thus in subjective analyses a STANDARD OBSERVER is defined as
the mean of an observer population. For example in a recent subjective scaling
of sharpness in print images, a subject is asked to place the image on a scale
referenced by a series of anchor images. While his response is variable we see
even in a population of eight observers that the observers exhibit the
characteristics of a normal distribution whose mean is well defined and stable.
Image rulers are a very simple way of obtaining subjective evaluations of image
properties. Next month we will discuss some of the uses of image rulers in
subjective measurements.
Please contact us to discuss how we can assist you in designing and measuring
subjective properties of imaging systems.
Newsletter Archives
Newsletter archive: The intention of these newsletters is both to inform
and promote our products and services. To this end we will maintain an archive
of past newsletters on our website: We invite you to visit and hope that the
information will be of value.
Bror Hultgren
Image Integration, Inc.
978.356.5975
bror.hultgren@i-2-q.com
ImagIntegration.com