Quality control is one of the most important
issues in the industry of steed sheet production.
Detection of surface defects allocates a high percent
of quality control process to itself. Nowadays in most
production lines of steel sheets, quality control is
executed manually by expert personnel. The lack of
an automated system for quality control causes a
decrease in efficiency, increases costs and makes
inaccuracy. Image Processing is dominant technology
today for inspection of different tissues and
recognition of available diversity. The power of this
technology, especially in two fields of detection and
classification of the template, makes it possible to
utilize in quality control of such industries as textile,
paper and ceramic. With regard to such practical
background of techniques for image processing and
the sort of surface defects available on the steel sheets,
lots of researches has been made so far for the
purpose of automatic detection of defects [1&2]. The
majority of attempts so far made, focused on the
features of color and figure in color to offer
appropriate methods, In [3] the feature of tissue has
been used, in which by utilizing Gabor wavelet for
signal processing by MSMD method, tissue features
of steel sheets are extracted. Since the availability of a
normal image is vital for defect detection and proper
selection of normal image is important (it must have
the same background as that of defective image ) and
the comparison of normal image with defective ones
is time –consuming and whit regard to the fact that
detection of partial image which must be deleted is
performed manually, in the new method the necessity
to normal image is removed and the quantity of partial
images to be deleted is determined automatically, so
that the speed of defect detection is increased, whit