Basic HTML Version
PerfectScan White Paper – Copyright © Dynamic Computing Solutions
2013
2
algorithms are tuned specifically to reduce noise, achieve perfect gamma correction, and obtain optimal
image clarity. The end result is an output image that was never before thought possible.
T
HE
S
CIENCE BEHIND PERFECTSCAN
®
Denoise
Symbol
Extraction
Symbol
Refinement
Symbol
Sharpening
Background
Removal
Global
Threshold
Cognitive Process
Input
Image
Output Image
PerfectScan
®
Figure 2: Processing steps of PerfectScan
®
technology
Denoise
. Both spatial and intensity information between a point and its neighboring points are
considered, unlike the conventional linear filtering where only spatial information is considered.
Edges and sharpness of the image are preserved.
Symbol Extraction
. Symbols or objects are analyzed and extracted from the image.
Symbol Refinement
. Disconnected symbols are intelligently repaired by connecting the missing
line segments.
Symbol Sharpening
. During this phase, the edges of the symbols are sharpened to eliminate the
noises near the edges, as well to reduce the pixilation effect.
Background Removal
. Background information is intensively dissected. The uniform
background is
dropped, whereas textured background is partially retained. In certain instances it
is desirable to retain the background. Therefore, this functionality is optional.
Global Threshold
. Adaptive thresholding is employed. The optimal number of threshold levels
is determined by computing the number of significant peaks from the image’s histogram.
Cognitive Merging Process
. This is the heart of PerfectScan®. All the intermediate results are
combined and analyzed using a set of cognitive-based algorithms. Optimization goals reduce
noise, achieve perfect gamma correction, and obtain optimal image clarity.
OCR
IMPROVEMENT
Current solutions to the problem of optical character recognition (OCR) have advanced to the point where
recognition rates of 99% are common for clean and uniformly formatted text. Unfortunately, the
performance of most OCR algorithms degrades very rapidly when even small amounts of noise are
introduced into the original document or during the scanning process. In many situations, this increased
error rate quickly decreases the ROI to the point where it is not cost-effective to integrate automated
recognition technology solutions.
With the unique character rebuilding functionality provided by PerfectScan
®
, the OCR result and the
readability of an image is dramatically improved. Additionally, PerfectScan
®
automatically changes its