Accelerating Computation of Zernike and Pseudo-Zernike Moments with a GPU Algorithm

Although Zernike and pseudo-Zernike moments have some advanced properties, the computation process is generally very time-consuming, which has limited their practical applications. To improve the computational efficiency of Zernike and pseudo-Zernike moments, in this research, we have explored the use of GPU to accelerate moments computation, and proposed a GPUaccelerated algorithm. The newly developed algorithm is implemented in Python and CUDA C++ with optimizations based on symmetric properties and k × k sub-region scheme. The experimental results are encouraging and have shown that our GPU-accelerated algorithm is able to compute Zernike moments up to order 700 for an image sized at 512 × 512 in 1.7 seconds and compute pseudo-Zernike moments in 3.1 seconds. We have also verified the accuracy of our GPU algorithm by performing image reconstructions from the higher orders of Zernike and pseudo-Zernike moments. For an image sized at 512 × 512, with the maximum order of 700 and k = 11, the PSNR (Peak Signal to Noise Ratio) values of its reconstructed versions from Zernike and pseudo-Zernike moments are 44.52 and 46.29 separately. We have performed image reconstructions from partial sets of Zernike and pseudo-Zernike moments with various order n and different repetition m. Experimental results of both Zernike and pseudo-Zernike moments show that the images reconstructed from the moments of lower and higher orders preserve the principle contents and details of the original image respectively, while moments of positive and negative m result in identical images. Lastly, we have proposed a set of feature vectors based on pseudo-Zernike moments for Chinese character recognition. Three different feature vectors are composed of different parts of four selected lower pseudo-Zernike moments. Experiments on a set of 6,762 Chinese characters show that this method performs well to recognize similar-shaped Chinese characters.

A MACH Filter-Based Reconstruction-Free Target Detector and Tracker for Compressive Sensing Cameras

Compressive sensing cameras hold the promise of cost-effective hardware, lower data rates, and improved video quality, particularly outside the visible spectrum. However, these improvements involve significant computational cost, as sensor output must be reconstructed in order to form an image viewable by a human. This paper describes a prototype automated detection and tracking system using a compressive sensing camera that does not rely on computationally costly image reconstructions. It operates on raw sensor data for an approximately ten-fold improvement in computation time over a comparable reconstruct-then-track algorithm. The detector is successful at a sensing rate of 0.3, comparable to that required for high-quality image reconstructions. If initialized with the location of a target, the tracker holds the target at a sensing rate of 0.005, below the boundary where reconstruction breaks down. These results show not only that direct tracking from compressive cameras is possible, but also give support to the pursuit of direct inference from compressive sensors of all types.

Autonomous Magnetic Resonance Imaging

Access to Magnetic Resonance Imaging (MRI) across developing countries from being prohibitive to scarcely available. For example, eleven countries in Africa have no scanners. One critical limitation is the absence of skilled manpower required for MRI usage. Some of these challenges can be mitigated using autonomous MRI (AMRI) operation. In this work, we demonstrate AMRI to simplify MRI workflow by separating the required intelligence and user interaction from the acquisition hardware. AMRI consists of three components: user node, cloud and scanner. The user node voice interacts with the user and presents the image reconstructions at the end of the AMRI exam. The cloud generates pulse sequences and performs image reconstructions while the scanner acquires the raw data. An AMRI exam is a custom brain screen protocol comprising of one T1-, T2- and T2*-weighted exams. A neural network is trained to incorporate Intelligent Slice Planning (ISP) at the start of the AMRI exam. A Look Up Table was designed to perform intelligent protocolling by optimising for contrast value while satisfying signal to noise ratio and acquisition time constraints. Data were acquired from four healthy volunteers for three experiments with different acquisition time constraints to demonstrate standard and self-administered AMRI. The source code is available online. AMRI achieved an average SNR of 22.86 ± 0.89 dB across all experiments with similar contrast. Experiment #3 (33.66% shorter table time than experiment #1) yielded a SNR of 21.84 ± 6.36 dB compared to 23.48 ± 7.95 dB for experiment #1. AMRI can potentially enable multiple scenarios to facilitate rapid prototyping and research and streamline radiological workflow. We believe we have demonstrated the first Autonomous MRI of the brain.