Multiplicative Noise Removal Of Color Images Based On High-order Model

In the process of transmitting and storing color digital image information,pattern noise is usually generated,and is mainly divided into additive noise and multiplicative noise.In recent years,with the research on additive noise gradually becoming mature,related scholars have gradually started to conduct extensive research on multiplicative noise,and proposed corresponding models to remove multiplicative noise for color images.However,as the processing of color images requires consideration of the coupling relationship between layers,traditional multiplicative noise removal models for color images can cause many problems when processing noisy images,such as step effects and blurred edges.In this study,two new models,SO-EE(SO-Euler-Elastica)and SO-TGV(SO-Total Generalized Variation),were proposed to address the above problems in the traditional noise removal models for color images.1.In the SO-EE(SO-Euler-Elastica)model,the Euler-Elastica term was used as the higher order rule term in the denoising model and combined with the generalised multiplicative noise data term in the SO model.During the solution process,the Split Bregman fast solution algorithm was devised and an alternating optimisation algorithm was iterated for each subproblem.The FFT(Fast Fourier Transform)algorithm and soft thresholding formulation were also introduced in the solution process.2.The SO-TGV(SO-Total Generalized Variation)model involves using the TGV term as a higher order regular term in the denoising model and combining it with the generalized multiplicative noise data term in the SO model.Furthermore,the Split Bregman fast solver algorithm and the alternating optimisation algorithm were designed to solve the model.In this study,the two models were applied to color image multiplicative noise removal.The two models were compared with the conventional CTV(Color Total Variation)model,the MTV(Multi-channel Total Variation)model and the higher order TC(Total Curvature)and TGV(Total Generalised Variation)models.At a later stage,qualitative and quantitative experiments were used to verify the denoising effect of the two models.The experimental results show that the images processed by the two new models maintain better sharpness,contrast,etc.,and avoid step effects and blurred edges.The new model achieves higher PSNR(Peak Signal-to-Noise Ratio)and SSIM(Structural Similarity Index Measurement System)values when processing noise than the conventional denoising model.Furthermore,the model takes less time to iterate through the noise process than other higher order models.