In subpixel shift super-resolution (SR) imaging, accurate sub-pixel image registration is a key issue. Traditional superresolution reconstruction methods use a motion estimation algorithm to estimate a shift, and then adopt different methods for SR image reconstruct. In this paper, we focus on designing a SR imaging system, in which instead of moving a camera only, the imaging lens before the camera is also moved. By doing so, we reduce the shifting resolution requirement. As the camera with the lens move 13μm, the image moves 1μm. A set of 16 or 9 low-resolution (LR) images of a scene are captured with the system. The sub-pixel shifts between these LR images are 1μm and 2μm, respectively. Then Projection onto Convex Sets (POCS) algorithm is used to reconstruct the SR image. The results show much higher spatial resolution comparing to the LR images.
In many situations, imagers are required to have higher imaging speed, such as gunpowder blasting analysis and observing high-speed biology phenomena. However, measuring high-speed video is a challenge to camera design, especially, in infrared spectrum. In this paper, we reconstruct a high-frame-rate video from compressive video measurements using temporal compressive imaging (TCI) with a temporal compression ratio T=8. This means that, 8 unique high-speed temporal frames will be obtained from a single compressive frame using a reconstruction algorithm. Equivalently, the video frame rates is increased by 8 times. Two methods, two-step iterative shrinkage/threshold (TwIST) algorithm and the Gaussian mixture model (GMM) method, are used for reconstruction. To reduce reconstruction time and memory usage, each frame of size 256×256 is divided into patches of size 8×8. The influence of different coded mask to reconstruction is discussed. The reconstruction qualities using TwIST and GMM are also compared.
Compressed Sensing (CS) can use the sparseness of a target to obtain its image with much less data than that defined by the Nyquist sampling theorem. In this paper, we study the hardware implementation of a block compression sensing system and its reconstruction algorithms. Different block sizes are used. Two algorithms, the orthogonal matching algorithm (OMP) and the full variation minimum algorithm (TV) are used to obtain good reconstructions. The influence of block size on reconstruction is also discussed.
As compressive imaging can capture high-resolution images using low-resolution detectors, it has received extensive attention recently. Compared to Single-pixel Compressive imaging, block compressive imaging (BCI) can considerably reduce the observation and calculation time of the reconstruction process, therefore it can also reduce the speed of imaging. A common challenge in BCI implementation is system calibration. In this paper, we use system spread point function into object reconstruction process to solve this challenge. In our simulation works, a 64x64 object with block size 4x4 is used. 6 measurements are collected for each block. Orthogonal matching pursuit (OMP) algorithm is applied to reconstruction. Additionally, we setup an experiment to demonstrate BCI idea. The BCI experimental platform confirms that images at high spatial resolution can be successfully recovered from low-resolution sensor.