Deep demosaicking convolution neural network and quantum wavelet transform-based image denoising.

Area of Science:

• Digital Image Processing
• Computational Imaging
• Machine Learning for Image Restoration

Background:

• Digital color images are captured using monochrome sensors with Color Filter Arrays (CFAs).
• Demosaicking and image denoising are crucial for restoring full-color images.
• Existing methods for combined image restoration are being actively researched.

Purpose of the Study:

• To develop a deep learning (DL) based model for simultaneous image denoising and demosaicking.
• To design an Autoregressive Circle Wave Optimization (ACWO) based Demosaicking Convolutional Neural Network (DMCNN) for demosaicking.
• To integrate Quantum Wavelet Transform (QWT) for enhanced image denoising.

Main Methods:

• A novel Demosaicking Convolutional Neural Network (DMCNN) with Autoregressive Circle Wave Optimization (ACWO) was designed.
• Quantum Wavelet Transform (QWT) was employed for image denoising, analyzing noise, and applying soft thresholding.
• The denoised and demosaicked images were fused using a weighted average technique.

Main Results:

• The QWT+DMCNN-ACWO model achieved high performance metrics: PSNR of 49.549 dB, SDME of 59.53 dB, SSIM of 0.963, and FOM of 0.890.
• The model demonstrated efficient computational time.
• The fused image effectively combined denoising and demosaicking results.

Conclusions:

• The proposed QWT+DMCNN-ACWO model offers a robust solution for simultaneous image denoising and demosaicking.
• This deep learning approach significantly improves image restoration quality.
• The model shows promise for advanced digital imaging applications.