Denoising Algorithms Based On Wavelet For Pools Underwater Image

With the improvement of living standard, swimming has become one of the dailyentertainments, especially in the hot summer; many people will like to choose swimming. Butthe life-saving system of the swimming pool is imperfect, so that the accident of drowningoften happens in swimming. This paper focuses on under-water image de-noising techniquefor swimming pool, lays the foundation for the follow-up work about swimming pool alarmsystem. Due to the water particles, dust, light and swimmers from the movement of waterripples, the signal-to-noise ratio is very low, the traditional image de-noising methods cannotget good results. For the image features in the pool, we have adopts the wavelet de-noisingalgorithm and main research the wavelet threshold de-noising algorithm, analysis of theadvantages and disadvantages of wavelet threshold algorithm and put forward the newadaptive wavelet threshold de-noising algorithm.In this paper, we first have carried on the detailed analysis of the properties ofunderwater images, light refraction and scattering effect of water to have serious impact onthe imaging process. After the traditional image de-noising algorithm is discussed, and pointsout its existing deficiencies and in need of improvement. In order to find a effectivede-noising algorithm about swimming pool underwater image, this paper introduces thewavelet transform based on Fourier transform, wavelet has the characteristics ofmulti-resolution analysis and Mallat algorithm, show its application to image de-noising isfeasible, and has the great advantage. Based on detailed analysis of the wavelet thresholdde-noising algorithm, the shortcoming of traditional threshold was improved. First, thewavelet decomposition level, the selection of Gaussian noise test model decomposed waveletcoefficients noise test to determine whether the next step decomposition, enhanced adaptivealgorithm; followed in threshold selection, the selection of Modified Bayesshrink thresholdformula calculation, so that the different layers corresponding to different thresholdprocessing, increasing the algorithm is reasonable; finally is the threshold function aspect, animproved non-linear threshold function and no-parameter threshold function, the former hasbetter flexibility, but need to optimize the parameters selected, the latter can be more completeimplementation of adaptive algorithms.Simulation experiments are implemented by MATLAB wavelet toolbox, through theexperiment results; shows that the new wavelet threshold de-noising algorithm usually obtainsbetter performance than others.