Wavelet Analysis of Image Denoising
Abstract: The wavelet analysis of the basic principles of de-noising, wavelet transform based
on the image denoising method and experimental verification. The results showed that with the
commonly used methods of image denoising, wavelet denoising method can retain the good image
of the details of information, performance is superior to other methods.
Key words: image noise, wavelet analysis, denoising
1 Introduction
Images in the collection, transmission and conversion in the external environment is often
subject to interference. Inclusion of the image noise and reverberation interference, not only made
the image quality fell, affecting the image of the visual effects, and to the image processing has
also brought further inconvenience. In order to reduce interference noise on the image to avoid
miscarriage of justice and missed to remove or reduce the noise it is necessary to work.
2 Commonly used methods of image denoising
Common method to remove noise in the neighborhood averaging method, filter method. Jane
introduced the following several methods of image denoising.
(1) Average Neighborhood
Neighborhood averaging is a partial treatment of space. -based image smoothing image
for , It's the gray value of each pixel included in the from the development of
a number of neighborhood pixel gray value of the average decision, will be subject to interference
model into a two-dimensional images with the airport, the general noise are additive, iid Gaussian
white noise. If the definition of signal to noise ratio for images with noise and the average ratio of
noise variance, then by smoothing the image with noise, its SNR will be increased times (
is adjacent to a domain that contains the number of pixels). It can be seen that the
neighborhood used an average radius of the larger neighborhood, the greater the signal to noise
ratio to improve, and after smoothing the image more blurred, the details of the distribution of
information was not obvious.
(2) Time domain frequency domain low-pass filtering method
For an image, its edge, jumping as well as image noise for high-frequency component, and the
background of a large area and slow-varying part of the district on behalf of low-frequency
component images, you can design a suitable low-pass filter to remove high-frequency component
to In addition to noise. Often used in time-domain image and low-pass convolution with
convolution of the template approach to image denoising, the main steps are:
(a) the template in the chart followed by the mobile, so that the template center and map the
location of a pixel overlap;
(b) the template with the template in the coefficient of the corresponding pixel multiplied;
(c) the sum of all product;
(d) the value assigned to coincide with the center template pixel point;
In particular, when the template for each value of 1, is the neighborhood average. For the case of
frequency domain, as shown in Figure 1.