Signal Processing for Stereoscopic and Multi-View 3D Displays - Signal Processing Systems - page 26

each view at some selected points and searching for single function that gives the
best fit for all measurements, regardless of the angle.
The first step of the measurement technique is to prepare two groups of test
images. The first group consists of so-called single view images where only the
sub-pixels from one view are lit. These images are used for measuring the angular
visibility. The second group contains test images where all pixels are set to different
levels of grey in order to linearize the camera response function [ 64 ] . In the second
step, each test image is shown on the test display and is photographed from a number
of observation positions. The observations positions are selected on a line parallel
to the display surface and at the optimal viewing distance. If the measurement
point is displaced from the center of a visibility zone, the visibility function gets
sampled with an offset and the maximum value of that function falls in between
two samples. However, judging by measurement results in other works [ 54 , 59 ,
65 ] , one can assume that the visibility function for all observation points can be
closely approximated by the same function, which has its peak occurring in the
optimal observation spot for the corresponding view. In the third step, based on this
assumption, one can search for single function that closely fits measurements for all
positions regardless of possible offset.
More details about measuring the angular visibility function of a 3D display can
be found in [ 35 , 54 , 59 , 60 , 63 ] .

3.2.4 Display Passband

Spatially-multiplexed 3D displays suffer from masking distortions and fixed-pattern
noise caused by visible gaps between the pixels and/or by apparent non-rectangular
shape of a pixel. The visibility of such distortions depends on interaction between
the spectrum of the visualized content and the display's transfer function. This
interaction can be conveniently expressed in the frequency domain. Therefore, in
order to assess the visibility of masking, one needs to study the performance of the
display in frequency domain through a quantity called a display passband .Inthis
subsection we present a simple yet efficient six-step methodology to measure the
display passband. The approach is shown in Fig. 20 .
The first step is to prepare a number of test signals which contain a 2D sinusoidal
pattern with varying horizontal and vertical frequency components, as the ones
shown in Fig. 21 a , c. Then, out of each test signal a number of test images, each
one with different apparent depth, are prepared. This is done by mapping the same
signal to each view of the display, adding different amount of disparity to each
view and interleaving all views in a test image. The third step involves automated
visualization of all test images on the display and making a snapshot of each one
with a high-resolution camera. The output of that step is a collection of test shots of
all test images, similar to the ones shown in Fig. 21 b , d. In the next step the spectrum
of each test shot is analyzed in order to determine the amplitude ratio between the
original frequency component in the test signal and the most noticeable distortion
frequency component introduced by the display.