Fixed Pattern Noise Removal Based on Wavelet Transform and Semi-Calibration

Non-uniformities in pixels are prevalent in existing image sensors. In low illumination environments, the pixel nonuniformities result in the undesirable fixed-pattern noise (FPN), which severely limits the imaging capability of image sensors in dark light environments. The FPN in each pixel can be considered as a linear scaled noise pattern with an uncertain scale parameter. Aiming at the image quality degradation caused by the FPN, a noise removal algorithm based on wavelet transform and semi-calibration is proposed. The key to this work is that since wavelet transform is a linear transform, the existing linear scaled FPN model is equally valid in wavelet domains. Therefore, the scale parameter estimation in spatial domains can be accomplished in wavelet domains. To implement the parameter estimation, this paper first establishes a parameter estimation framework for minimizing Bayes risk in spatial domains. Then, combined with the noise model in wavelet domains, this paper derives the parameter estimation based on the Bayes risk minimization in wavelet domains, which is implemented by the proposed alternating iteration algorithm. With the scale parameter and the semi-calibrated noise pattern, the FPN removal from the noisy image is ultimately accomplished. The proposed method is validated in simulation experiments on BSD500 dataset as well as real-world noisy images experiments. The simulation experiment results indicate that the proposed method shows an obvious enhancement in PSNR and SSIM evaluation metrics compared with the existing algorithms. Moreover, the processing speed of the proposed method is significantly faster than the existing semi-calibration method. Finally, this paper verifies the feasibility of the proposed method by implementing real denoising experiments.