Adaptive Wiener filter super-resolution of color filter array images

Area of Science:

• Digital Imaging and Signal Processing
• Computer Vision
• Image Reconstruction

Background:

• Digital color cameras with Bayer Color Filter Arrays (CFAs) rely on demosaicing to reconstruct full-color images from single detector arrays.
• Standard demosaicing methods do not adequately address inherent undersampling and aliasing issues in camera designs.
• Super-resolution (SR) techniques, particularly fast non-uniform interpolation-based methods, offer potential for aliasing reduction and real-time applications.

Purpose of the Study:

• To develop a novel fast super-resolution (SR) method for CFA cameras by adapting the adaptive Wiener filter (AWF) algorithm for color demosaicing.
• To investigate the application of the AWF SR algorithm, initially designed for grayscale imaging, to color SR demosaicing.
• To evaluate the performance of the proposed color AWF SR method, both as a standalone algorithm and as an initialization for variational SR algorithms.

Main Methods:

• Development of a novel fast SR method based on the adaptive Wiener filter (AWF) algorithm tailored for CFA cameras.
• Utilization of global channel-to-channel statistical models within the AWF SR framework.
• Application of the developed color AWF SR method as a standalone algorithm and as an initialization for a variational SR algorithm.

Main Results:

• Successful adaptation of the adaptive Wiener filter (AWF) SR algorithm for color super-resolution demosaicing in CFA cameras.
• Demonstration of the method's capability to reduce or eliminate aliasing inherent in typical camera designs.
• Performance comparisons with other SR techniques using both simulated and real image data, highlighting the effectiveness of the proposed approach.

Conclusions:

• The developed color AWF SR approach provides a computationally efficient and effective solution for improving full-color image quality from CFA cameras.
• The method addresses fundamental undersampling and aliasing issues more effectively than traditional demosaicing techniques.
• The proposed AWF SR algorithm shows promise for real-time applications and serves as a valuable initialization for more complex SR algorithms.