Image enhancement device and method for high-resolution video denoising

Edge-Preserving Image Denoising Using Perturbation Matrix-Based Wavelet Wiener Filtering

This technology is an image enhancement device and method that restores denoised images by decomposing noisy images into high-frequency and low-frequency signals via wavelet transform, creating expansion matrices for the high-frequency image and its orthogonal perturbation Y, estimating image variance from their elements, and calculating wavelet coefficients.

Wavelet-domain Wiener filters require estimating the original signal's variance from noisy data, which leads to inaccurate estimates in high-noise environments. Increasing the local window size to stabilize estimation often results in blurred details in edge regions.

This technology uses the MMSE technique to estimate perturbation matrix elements as a linear combination of four neighboring values and determines signal variance using high-resolution frequency information. This allows for the reduction of additive white Gaussian noise while preserving edges and textures, even with large windows. It can be applied to high-resolution broadcast post-processing and display image engines to reproduce sharp, clear content even when affected by low-light or transmission noise.

‍

‍

Key Features:

• Receiving a noisy image X and decomposing it into high-frequency and low-frequency signals via wavelet transform
• Generating an expanded matrix for the high-frequency noisy image X and an orthogonal perturbation Y of the same size
• Estimating image variance based on the expanded matrix elements to calculate wavelet coefficients, then performing element-wise multiplication and summation to reconstruct the original image
• A perturbation estimator that minimizes the mean square error in the linear combination of four neighboring elements of perturbation Y using the MMSE technique

‍