Image Denoising Using Multi-Frame Image Content in Curvelet Transform Domain

We present a multi-frame information involved image denoising approach through spatiotemporal GSM (Gaussian Mixture Scales Modeling) of image Curvelet statistics both considering of the innerframe and inter-frame statistics. By using the Bayes inference based least square estimation, we construct a Bayesian Least Squared GSM (BLS-GSM) based image denoising model for single image noise cancellation and estimate the optimal coefficient of the uncontaminated image through this model in the Curvelet domain in advance. Then, we carry out a novel spatial-temporal jointed image noise removing method by combining the single image denoising model with a weighted impact factor conducted on multi-frame images based on the relativity of the image coefficients among the image sequences. The image denoising performance is evaluated in terms of the visual effect of the denoised images and compared against the other methods by the objective criterion, Peak Signal-to-Noise Ratio. The comparative results demonstrated that this method outperforms the Curvelet based counterpart and achieves close or better processing results in terms of PSNR, namely that the proposed method is capable of achieving higher reconstruction quality while protecting more image details.