Research On Image Filtering Method Based On MMTD And Wavelet Transform

Image filtering is a process of reducing the noise of image while keeping image’s detail. Theeffect of image filtering plays a key role in the subsequent phase of image processing andanalyzing.In this paper, the image filtering algorithms based on spatial filtering, frequency filtering andwavelet domain filtering are studied and compared. Since the traditional wavelet hard thresholdfiltering algorithm has the shortcomings of discontinuousness, a MMTD-based wavelet hardthreshold image filtering algorithm is proposed by using MMTD to determine the value of waveletcoefficient based on the fuzziness of wavelet coefficient around the threshold. Experimental resultsshow that the new algorithm has better filtering effect than traditional ones when dealing withstrong noise.Several improved algorithms of threshold function are also studied in this paper and a newBayes soft threshold image filtering algorithm combined MMTD and wavelet correlation isproposed. The proposed algorithm divides the value range of wavelet coefficient into three partsaccording to the ratio function of distance in MMTD. The part where the value of waveletcoefficient is small is processed by wavelet correlation algorithm which could retain the imagedetail mistakenly filtered before. The part where the value of wavelet coefficient is medium isprocessed by MMTD-based wavelet hard threshold image filtering algorithm and the value ofwavelet coefficient is calculated by atrophying as the soft threshold image filtering does. The partwhere the value of wavelet coefficient is large is processed through atrophying the value of waveletcoefficient by using the ratio of threshold to wavelet coefficient. Experimental results show that thenew algorithm keeps more detail and has better filtering effect than traditional one.