Paper instructions:
A digital image is an image (x,y) with a discrete value to both its
spatial coordinates and brightness. Its representation is in two
dimensional integer array, series of 2D arrays or one for each colour
band with a digitised brightness value called grey level. In the integer
array, pixel or pel is the name given to each of the elements. This name
is derived from ‘picture element’ (Maria and Costas, 2010). Image
processing is a response to problems associated with images in
digitisation and coding for transmission, printing and storage purposes,
image enhancement and restoration, and image segmentation and
Generally, image processing encompasses all forms of signal processing
when image is its input, and the output being an image or image-related
parameters. The image processing processes are essentially done for
different purposes using different befitting techniques. Some of the
popular images processing processes are: visual processing 2D and 3D
modelling, image rendering, computational photographs and so on (Maria
and Costas, 2010; Szeliski, 2010; Buhmann et al., 1999). Present
experiences have necessitated the need for works related to processing
and majorly recognizing digital images for varieties of real life
applications: ranging from biometric, medical imaging, robotic, remote
sensing, forensics and security. Digital images as set of sampled data
mapped onto a grid of colour in 2-dimension (2D) or represented using
digits and pixels have also attracted diverse scientific researches (Kuo
et al. 2006).

According to Szeliski (2010), among the computer vision tasks,
recognising image objects in respect to their constituents is the most
challenging. Recognition is said to be hard because of the composing
objects that made up of the image objects, extreme variation in shapes
and sizes, difference in appearance and the non-rigid articulation. These
are few of the reasons why image recognition is challenging as an
industrial task, and also, as an academic pursuit. This probably account
for the low quantity of image recognition’s work when compared to other
computer vision specialities.
Issues in image recognition can be disintegrated into different sections:
Object detection is when what is being looked for is known, instance
recognition is a situation whereby a specific rigid object is to be
recognised, and class recognition is recognition of a general category of
objects. Szeliski (2010) opined that: object detection involves general
scanning of the image to determine its matching characteristics, the
characteristic feature points and the verification of its alignment in a
geometrical way is instance recognition’s responsibility, while
recognizing that extremely varied class’ instances is involved in class
recognition. The recognition of the Arabic characters can safely be
categorised under object detection. In this case, although it is in
digital form, the need for matching characteristics with the feature
extraction is similar.
As Nieddu and Patrizi (2000) observed, pattern recognition is a
simulation of mathematical, statistical and heuristic techniques for
accomplishment of qualitative performance of human capacity. It has
helped in concretizing machine learning and pattern detection in computer
implementation with the assistance of artificial intelligence (AI) and
computer vision. Considering the wide range of pattern recognition
application, Selim (2012) pointed out that: multimedia database retrieval
through internet search is one of the applicable areas of pattern
Therefore, Selim’s (2012) assertion of the domain applicability of
pattern recognition qualifies to attend to this domain problem. However,
with extension to content-based image retrieval (CBIR) since online
Arabic character image recognition without a textual description is under
consideration. In this case, browsing, searching and retrieving images
from a large database of digital images are the broad area of this study.
The online Arabic character recognition is the intersection between
‘search’ and ‘retrieve’ processes using content similarity such as
texture, colour and shape. This essentially is an application of
computer vision to image retrieval (Szeliski, 2010). Figure 1 depicts the
CBIR process and the image recognition phase
Additional materials: not defined