01-01-2013, 11:22 AM
Correspondence
Correspondence.pdf (Size: 242.32 KB / Downloads: 17)
Abstract
To assess the performance of image quality metrics (IQMs),
some regressions, such as logistic regression and polynomial regression, are
used to correlate objective ratings with subjective scores. However, some
defects in optimality are shown in these regressions. In this correspondence,
monotonic regression (MR) is found to be an effective correlation method
in the performance assessment of IQMs. Both theoretical analysis and experimental
results have proven that MR performs better than any other
regression. We believe that MR could be an effective tool for performance
assessment in the IQM research.
INTRODUCTION
To assess the performance of image quality metrics (IQMs), a
scheme first proposed by Video Quality Expert Group (VQEG) [1]
is widely adopted by researchers. The scheme is designed for the
objective measurement evaluation problem and can be applied for the
assessment of image/video quality metric. It can be described by the
following three steps:
Step 1. Metric computation
Rate a set of images by IQM, while mean opinion scores (MOSs)
of these images are measured by human observers beforehand.
Step 2. Correlation (or regression)
Correlate the outputs of IQM with MOS via a predicting function
(this process can be also called regression). Then, the predicted
MOS is obtained by calculating the IQM ratings through the regression
function.
Step 3. Index computation
Compute the performance indexes between the MOS and the predicted
MOS.
CORRELATION IN PERFORMANCE ASSESSMENT OF IQM
VQEG’s description of the correlation method [2] is as follows.
As the nature of the nonlinearities was not well known beforehand,
it was decided that two different functional forms would be regressed
for each model, and the one with the best fit (in a least-square sense)
would be used for that model. The functional forms used were a thirdorder
polynomial and a four-parameter logistic function. The regressions
were performed with the constraint that the functions remain
monotonic over the full range of the data.
MR FOR CORRELATION
VQEG emphasize the monotonicity of regression, which guarantees
the monotonic relationship between metric ratings and human perception.
In other words, any monotonic function