The goal for this project is to implement an image processing pipeline in MATLAB with the addition of a superresolution algorithm from literature (Daniel Glasner).
The pipeline begins with a defective pixel correction module based on median filtering. Followed by black-level subtraction, lens shading correction (LSC), and white balance (WB) modules. The LSC module applies gain LUT on the input image. The LUT is for the same illuminant as the input image (for better color shading correction), applied per Bayer color channel. WB is implemented using the shades-of-gray algorithm.
The next block to be implemented is Bayer denoising (Dmitriy Paliya). This method proposes joint denoising and demosaicing and, according to the authors, “performs, at a lower computational cost, better or comparable than combination of successive state-of-the-art techniques targeted denoising and demosaicing techniques known to the authors.”
After demosaicing, a color correction matrix (CCM) is applied followed by gamma-correction (GMC), sharpening (Johns Hopkins University, 2005; Stanford University) and finally, superresolution. The CCM is a pre-computed 3×3 transformation matrix for the specific illuminant in the scene. GMC is a simple LUT of values.
At this stage, the superresolution algorithm is implemented and images from various scenes are simulated in the pipeline. The output is going to be analyzed and adjustments in the pipeline will be made accordingly.