Keywords
Image quality evaluation, image processing, quality
metrics, QoE.
1. Introduction
In the first part of this paper [1] the authors presented
a set of MATLAB-based applications useful for image
processing and image quality assessment developed at the
Multimedia Technology Group (MMTG), Faculty of Electrical
Engineering (FEE), Czech Technical University
(CTU) in Prague. These are the Image Processing Application,
the Image Quality Adjustment Application, the Image
Quality Assessment Application, the Image Quality
Evaluation Applications and the Results Processing Application.
All of them contain user-friendly interface which
make the usage intuitive and easy even for the users not
particularly educated in the field of image processing.
Related works to the applications were referenced also
in the Part I [2] – [12].
The purpose of this paper is to show some typical
examples from this field where the applications can be
useful. It means that the experimental results in this paper
are just a selection from the wide range of possible use.
The paper is organized as follows. Section 2 contains
the examples obtained from the Image Processing Application.
Section 3 shows the possible use of Image Quality
Adjustment Application. The illustration of subjective
quality assessment using the Image Quality Evaluation
Applications and following processing of its results with
Results Processing Application is in sections 4 and 5. Section
6 is about Image Quality Assessment Application and
section 7 concludes the paper and discusses the future
work.
2. Image Processing Application
The Image Processing Application is the largest one
and therefore offers the greatest number of image modifications.
It is divided into five subunits.
The first subunit is called Intensity Transformations.
It enables to simulate intensity distortions on one hand and
to increase the contrast and with that also the image quality
on the other. The examples of use are in Fig. 1. Fig. 1a) is
the original image, Fig. 1b) is the distorted version when
the low (high) output intensity thresholds were set higher
(lower) than low (high) input intensity thresholds (input
thresholds: low intensity – 0, high intensity – 1, output
thresholds: low intensity – 0.3, high intensity – 0.7). That
means that the dynamic range of the image was decreased.
Fig. 1c) represents the opposite case (input thresholds: LI –
0.15, HI – 0.85, output thresholds: LI – 0, HI – 1) where
the dynamic range was increased. Another important
characteristic of the picture that can be adjusted is the
gamma parameter. Fig. 1d) shows the impact on the original
image when the gamma parameter is set to be the half
of its original value. Fig. 1e) is the picture with gamma
equaled 1.5 times original value. This subunit also enables
to adjust the histogram of the image. It can either be
replaced by the histogram of the image uploaded by user or
modeled by bimodal gaussian function. An example of
modeled histogram is in Fig. 2. The picture with this histogram
is in Fig. 1f).
The second subunit of IP Application is the Spatial
Domain Filtering. It offers two main options – Linear and
Non-linear filtering. Linear filtering section contains number
of linear filters (complete list can be found in Part I of
this paper) based on fspecial( ) and imfilter( ) functions in
MATLAB. Results of this filtering are well known and
need no introduction