Cellular Automata for Image Resizing - Cellular Automata in Image Processing and Geometry - page 27

results is in many cases comparable or better than established specialised algo-
rithms. On the other hand, CA have been extensively used as a VLSI architec-
ture [34]. In contrast to the serial computers, the implementation of the model is
motivated by parallelism, an inherent feature of CA that contributes to further ac-
celeration of the model operation. The CA approach is consistent with the modern
notion of unified space-time. In computer science, space and time correspond to
memory and processing unit, respectively. In CA, memory (CA cell state) and pro-
cessing unit (CA local rule) are inseparably related to a CA cell [34]. In terms of
circuit design and layout, ease of mask generation, silicon-area utilization, and max-
imization of clock speed, CA are perhaps one of the most suitable computational
structures for hardware realization. To combine the inherent advantages of both CA
models and their implementation on hardware, when applied to image processing,
some researchers such as Andreadis et al. [2] proposed a Application Specific Inte-
grated Circuit (ASIC), which performs the conversion, in real time, of the R, G and B
colour co-ordinates to the CIE standard L*, a* and b* colour coordinates to be used
in colorimetry instrumentation for colour measurement and control and in colour
machine vision in autonomous applications such as robotics and military systems,
where the need for short processing times is crucial. In [3] a hardware CA mod-
ule for detecting circular objects is introduced, where targeted applications include
inspection tasks (accept/reject operations) of circular objects, such as tablets in the
pharmaceutical industry, and detection of uncoated areas, foreign objects and level
of bake in the confectionery and food industry. Moreover, Karafyllidis et al. [16]
calculate the mean velocity of a moving object, using the CA properties, along the
centra axis perpendicular to the lens of the vision system where the motion of the
object is restricted to translation (angular velocity is zero) and to one moving object
in the scene. It should be also mentioned that cellular logic image processors, like
Clip have been developed in the past and cellular neural networks, an extension of
CA that includes weight matrices, chips with both continuous time and discrete time
versions have been applied to a variety of image processing tasks [30]. In this as-
pect novel nanoelectronic structures like Quantum Cellular Automata (QCA) have
been also applied for image processing mathematical morphology operations [5, 25]
chapter 4. Finally, Porter et al. [27] proposed a CA reconfigurable framework which
includes a highly pipelined architecture for multi-scale cellular image processing
as well as support for several different pattern recognition applications, while Katis
and Sirakoulis [17] designed specialized FPGAs that achieve automated image pro-
cessing such as noise filtering, edge thinning and convex hull detection with the help
of corresponding CA algorithms.
Although it is clear that there many applications of CA to different image pro-
cessing tasks, and that while some such as image denoising are quite well known,
the application of CA to image resizing has not been explored in detail. In this chap-
ter, we present an edge-directed method which exploits the simplicity and the inherit
parallelism of the CA. The edges of a low resolution image are initially determined
by applying the Canny edge detector leading to a bit-wise edge map. This map
is then considered as a CA grid along with a CA state which corresponds to the
undefined pixel. The CA evolves its state by applying the appropriate CA rules