Towards the use of super-resolution in biomedical systems-on-chip

Abstract

Super-resolution is a smart process capable of generating images with a higher resolution than the resolution of the sensor used to acquire the images. Due to this reason, it has acquired a significant relevance within the medical community during the last years, especially for those specialties closely related with the medical imaging field. However, the super-resolution algorithms used in this field are normally extremely complex and thus, they tend to be slow and difficult to be implemented in hardware. This paper proposes a new super-resolution algorithm for video sequences that, while maintaining excellent levels in the objective and subjective visual quality of the processed images, presents a reduced computational cost due to its non-iterative nature and the use of fast motion estimation techniques. Additionally, the algorithm has been successfully implemented in a low-cost hardware platform, which guarantees the viability of the proposed solution for real-time biomedical systems-on-chip.