Comparative Study on Infrared Image De-noising Based on Matlab - Intelligent Computing for Sustainable Energy and Environment - page 73

Fig. 6. The results decomposed by one-level wavelet transform

that not only the de-noised effect is perfect, but the contrast of the infrared
image of transmission joints is enhanced by comparatively, which makes it easy
to judge the faults. This algorithm is ideal for Gaussian and multiplicative noise,
while it has little effect on salt& pepper noise.

S \

Fig. 7. The results de-noised by wavelet threshold combined with histogram equaliza-
tion

3 Comparison and Analysis of Simulation Results

In this image simulation processing study, Different de- noising algorithms are
taken to eliminate the Gaussian, salt& pepper and multiplicative noise respec-
tively. Comparing the final images using different methods of de-noising, we can
find out the suitable de-noising method for the specific noise.
Comparing Fig. 1 with (a), (b) of Fig. 8, we know that the de-noising effects
for Gaussian noise are not well whichever method is taken. Fig. 2 and (c), (d)
of Fig. 8 illustrate that the median filter is ideal for salt& pepper noise, though
some details of the image which have no effect on judging the faults are removed
too, so it can be for salt & pepper noise. Fig. 3 and (e), (f) of Fig. 8 shows that
the de-noising results are all not ideal by the three traditional methods, but if
some images need to be blurred, and then the low-pass filters would display its
advantages. For example, an image of character needs to smooth the local details,
and the low- pass filters can work well.Integrated analysis of Fig. 4 proves that
the multi-image average de-noising algorithm could remove all the three noises
and if possible, the more the average times is, the better the results will be.
Fig. 7 shows wavelet threshold united with histogram equalization method not
only removes the Gaussian and multiplicative noises perfectly but enhances the
contrast of results which makes it much easy to find the faults. But it has poor
effect on salt& pepper noise, because the high-frequency components of the noise