Comparison of Training Strategies for Autoencoder-Based Monochromatic Image Denoising.

Monochromatic images are used mainly in cases where the intensity of the received signal is examined. The identification of the observed objects as well as the estimation of intensity emitted by them depends largely on the precision of light measurement in image pixels. Unfortunately, this type of imaging is often affected by noise, which significantly degrades the quality of the results. In order to reduce it, numerous deterministic algorithms are used, with Non-Local-Means and Block-Matching-3D being the most widespread and treated as the reference point of the current state-of-the-art. Our article focuses on the utilization of machine learning (ML) for the denoising of monochromatic images in multiple data availability scenarios, including those with no access to noise-free data. For this purpose, a simple autoencoder architecture was chosen and checked for various training approaches on two large and widely used image datasets: MNIST and CIFAR-10. The results show that the method of training as well as architecture and the similarity of images within the image dataset significantly affect the ML-based denoising. However, even without access to any clear data, the performance of such algorithms is frequently well above the current state-of-the-art; therefore, they should be considered for monochromatic image denoising.

Multi-Wavelength Biometric Acquisition System Utilizing Finger Vasculature NIR Imaging.

Personal identification using analysis of the internal and external characteristics of the human finger is currently an intensively developed topic. The work in this field concerns new methods of feature extraction and image analysis, mainly using modern artificial intelligence algorithms. However, the quality of the data and the way in which it is obtained determines equally the effectiveness of identification. In this article, we present a novel device for extracting vision data from the internal as well as external structures of the human finger. We use spatially selective backlight consisting of NIR diodes of three wavelengths. The fast image acquisition allows for insight into the pulse waveform. Thanks to the external illuminator, images of the skin folds of the finger are acquired as well. This rich collection of images is expected to significantly enhance identification capabilities using existing and future classic and AI-based computer vision techniques. Sample data from our device, before and after data processing, have been shared in a publicly available database.

A Contemporary Design Process for Single-Phase Voltage Source Inverter Control Systems.

This paper presents an overview of contemporary voltage source inverter control system design. Design begins with the theoretical considerations that lead to the creation of the system's differential control law. This stage does not include scaling coefficients for the output voltage, output current, and filter inductor current. Following this, the inverter is modelled in MATLAB's Simulink environment with an appropriate load and control system. If the resultant simulation provides satisfactory results, a hybrid system consisting of MATLAB's Simulink and dSpace libraries with the MicroLabBox device is used to interface the simulation with an experimental hardware model in real-time. This allows the hardware plant and measuring traces to be validated. ControlDesk is used to scale the relevant coefficients. During the final stage of the design process, a microprocessor is programmed to control the inverter according to the dSpace simulation results. This requires new scaling values. Throughout every stage of the design process, too high a value of the modulation index disables the reduction of output voltage distortions. This paper details the entire design process for both single-input and multi-input control systems, explaining the scaling process and the required software. Such a modern design process ensures the shortest time between conceptualization and the final product.

Modified Distance Transformation for ImageEnhancement in NIR Imaging of Finger Vein System.

Most of the current image processing methods used in the near-infrared imaging of fingervascular system concentrate on the extraction of internal structures (veins). In this paper, we proposea novel approach which allows to enhance both internal and external features of a finger. The methodis based on the Distance Transformation and allows for selective extraction of physiological structuresfrom an observed finger. We evaluate the impact of its parameters on the effectiveness of the alreadyestablished processing pipeline used for biometric identification. The new method was comparedwith five state-of-the-art approaches to features extraction (position-gray-profile-curve-PGPGC,maximum curvature points in image profiles-MC, Niblack image adaptive thresholding-NAT,repeated dark line tracking-RDLT, and wide line detector-WD) on the GustoDB database of imagesobtained in a wide range of NIR wavelengths (730-950 nm). The results indicate a clear superiorityof the proposed approach over the remaining alternatives. The method managed to reach over 90%identification accuracy for all analyzed datasets.