Single-Pixel Detection Computational Imaging With Super-Resolution Based On Dictionary Learning

In recent years,single-pixel detection computational imaging has become a new research direction in the field of optical imaging relying on its novel imaging mechanism,which is based on compressive sensing theory.It uses a single-pixel detector to collect the linear measurements of the light in the scene,and reconstructs the image according to the correlative characteristics of the modulat ed light.Compared with traditional imaging,single-pixel detection computational imaging has a simple system,extensive range of applications,and low costs.In fact,in order to obtain higher resolution images,it utilizes the temporal resolution in exchange for the spatial resolution,which results a longer sampling process.Aiming at this problem,we present a method of single-pixel detection computational imaging with super-resolution based on dictionary learning,and research on the contents of several parts.In order to obtain a high-resolution reconstructed image,we propose an double-layer image super-resolution reconstruction algorithm based on feature classification.According to the detail of image in different layers is diverse,we use the dictionary sets obtained from different feature extraction methods to restore the details of various characteristics in the framework of double-layer reconstruction.And in the reconstruction process,a partial model of solving the sparse representation coefficients is adopted to make the reconstruction process more efficient.At the same time,the quality of the final image is optimized by local and non-local composite constraints.Simulation results show that the algorithm has a certain improvement in visual effect and objective measurement indicators compared with other classical algorithms.The traditional parallel sampling is converted to serial sampling in the existing single-pixel imaging model.In this paper,a system architecture of single-pixel detection computational imaging with super-resolution is proposed aiming at the fault of longer sampling time.After analyzing the function of each module,we give an improved experimental system using optical fiber to collect the modulated light,which eliminates the interference of the background noise caused by lens collection.Based on this,we design a super-resolution single-pixel imaging scheme combining the super-resolution algorithm in image reconstruction.The experiments prove that the scheme greatly inhibits the interference of external noise and reduces the time consumption,it also improves the quality of images.