Correction of intensity nonuniformity in breast MR images - Emerging Trends in Image Processing, Computer Vision, and Pattern Recognition - page 178

2 Preprocessing steps
The main purpose of the preprocessing stage is to enhance the image quality for further pro-
cessing steps by reducing or correcting the unrelated artifacts in the mammogram images.
Mammograms image analysis is problematic due to existence of different artifacts that make
the processing step complicated. Therefore, preprocessing stage is essential to improve the im-
age quality. It will prepare the mammogram image for the next processing stages.
2.1 Noise Reduction
Magnetic resonance images are corrupted by Rician distribution that arises from complex
Gaussian noise in the original frequency domain measurements.
The Rician probability density function for the corrupted MR image intensity x is demon-
strated as follows:
(1)
where σ is the standard deviation of the noise, A is the underlying true intensity, and I 0 is
the modified Bessel function of the first kind with order zero [ 15 ] . Median filter is one of the
most popular nonlinear spatial filters for noise reduction that is more efficient than convolu-
tion when the purpose is to preserve borders and decrease noise simultaneously. This method
is simple, computationally efficient, and also has a well denoizing power.
The median filter replaces the value of a given pixel with the median pixel value within a
region of interest. A median filter with properly selected window size can smooth the noise in
the original image. It may also virtually eliminate the main tissues information from the MR
image. Therefore, there will be a trade-off between noise reduction and the preservation of in-
formation from image. Clearly, by increasing the size of the median window, both noise sig-
nals and signals from main tissues are being suppressed. For the 7 × 7 case, it removed more
of the useful information than the 5 × 5 case. Therefore, the 5 × 5 median is the best for noise
removal and preservation of brain tissue information [ 16 ] .
The operation of median filtering technique is presented as:
(2)
Let Sxy and median be the set of coordinates in a subimage window which is centered at
( x , y ) and the median value of the window, respectively.