Kernel design for real-time denoising implementation in low-resolution images

Upsampling and removing noise from digital images are important tasks in image processing. Single-image upsampling with denoising influences the quality of the resulting images. Image upsampling is known as superresolution, which refers to restoration of a higher-resolution image from a given low-resolution image. In this paper, we propose a filter-based image upsampling and denoising method for low-resolution images. The proposed method involves two stages. In the first stage, we design least squares method-based filters. In the second stage, we implement an image upsampling and denoising process. The proposed method is compared with several standard benchmark methods, including the nearest neighbor, bilinear, and bicubic methods, to test whether it yields better restoration quality and computational advantages. In addition, we design various-sized filters and test them on low-resolution noisy images. From the experimental results, we conclude that filters with more taps return better results, but longer computational running times. The quality of the image upsampling and denoising of the tested methods is compared subjectively and objectively through simulation. The simulation results suggest how the user can best select an appropriate filter size to achieve optimal trade-off results.