FPGA implementation of linear combination for single pixel image reconstruction

Abstract

In this paper, an architecture for Single Pixel Imaging (SPI) reconstruction, the linear combination of a dictionary, and a measured sparse signal were implemented into a Field Programmable Gate Array (FPGA). The implementation is intended to be used to reconstruct images buried in high scattering media like biological tissue. A Walsh-Hadamard matrix generator was implemented as a dictionary to save memory blocks. Some tests were performed to evaluate its performance. These tests include the reconstruction of images at different noise levels and the reconstruction of images from sparse signals. The implementation was able to reconstruct a 128 by 128 pixels image with a reconstruction time just below 10 ms with significant computational time reduction. This reconstruction time could create animations and even 32 frames per second (FPS) video provided the acquisition speed is fast enough.