A Study On Image Adaptive Threshold Denoising Based On Dyadic Wavelet Transform

Image denoising is not only a hotspot in the discipline of image processing, but also a challenging research direction. In recent years, with the development of wavelet theory, wavelet transform according to its excellent properties has played a vital role in the field of image denoising. At present, the denoising method based on wavelet transform has become a major branch in image denoising and restoration, wavelet threshold denoising method which proposed by Donoho et with its simple principle, easy to implement has become one of the most active research. However, the wavelet threshold denoising method has many inherent defects, such as hard threshold function is not continuous, soft threshold function brings constant deviation between the estimated values and actual values, threshold short of scale adaptive and so on, these also limit its further application. On this basis, many researchers at home and abroad, from constructing new threshold functions and finding the optimal threshold made a lot of research. these opened up broad prospects for fully play its advantages as well as provided a basis for exploration of adaptive denoisng method. Taking an existing adaptive threshold image denoising, mostly study on orthogonal Wavelet transform, but for adaptive threshold image denoising research based on dyadic wavelet transform is truly rare.To search for more efficient denoising method and further enrich the content of wavelet denoising, basing on wavelet threshold denoising method, judging by peak signal to noise ratio and visual effect, from the algorithm, threshold function and threshold point of view, by means of experimental comparative analysis, this thesis makes exploratory study on adaptive image denoising method. Mainly includes the following:(1) Make a systematic exposition about Wavelet transform basic theory, emphatically introduce the discrete wavelet transform fast algorithm: Mallat algorithm and the dyadic wavelet transform fast algorithm:àtrous algorithm;(2) About problems of wavelet threshold denoising method makes a deep analysis and discussion, such as the choice of threshold function, determination of threshold, wavelet basis selection and determination of wavelet decomposition level;(3) Basing on advantages and disadvantages of hard and soft threshold function, taking Gaussian noise as background, we construct a kind of adjustable parameters adaptive threshold function, simulation results show that the function is better than soft threshold function and hard threshold function;(4) To inhibit Gibbs visual distortion which appears in denoised image caused by the lack of translation invariance in orthogonal wavelet transform, we introduces a translation invariant dyadic wavelet transform, and for the above experimental results, propose a kind of adaptive threshold denoising method based on dyadic wavelet transform. For the defect of Donoho universal threshold does not have the scale adaptive, resulting in loss of image detail, the method cites a kind of scale adaptive threshold. Simulation results show that not only in terms of peak signal to noise ratio but aslo in the visual effect, adaptive threshold denoising method based on dyadic wavelet transform has significantly increased and improved than wavelet hard threshold denoising method and soft threshold denoising method. It also shows that adaptive threshold denoising method based on dyadic wavelet transform proposed in this article is an effective method for image denoising.