Academia.edu uses cookies to personalize content, tailor ads and improve the user experience.
By using our site, you agree to our collection of information through the use of cookies.
To learn more, view our Privacy Policy.×
2002 11th European Signal Processing Conference, 2002
A novel circuit for a simple non motion-compensated deinterlacer is presented, where a motion det... more A novel circuit for a simple non motion-compensated deinterlacer is presented, where a motion detection module effectively computes a weighted mean of a temporal and a spatial interpolator. Both the motion detection module and the two interpolators are innovative, simple and effective, so that they are able to provide very good results, as experimental results show, at a low computational complexity. The final circuit, based on a pipeline architecture, has been designed to work on a real time video digital signal with a low power consumption.
Proceedings of the Fourth International Conference on Computer Vision Theory and Applications
We present techniques for the amplification of small contrast of bounded signals; one is based on... more We present techniques for the amplification of small contrast of bounded signals; one is based on gamma correction and another is of an unsharp-masking type; the one of the unsharp-masking type is suitably modified for its application on circular signals as well. We enhance the saturation and luminance components of high dynamic range images on the basis of a segmentation of the image into light and dark regions.
Adaptive structure based on neural networks for subband-filteredimage compression
A nonlinear pseudo-Retinex for dynamic range compression in CMOS imaging systems
Sviluppo di un algoritmo per il miglioramento di immagini acquisite in condizioni ambientali sfavorevoli
An FPGA Realization for Real-Time Depth Estimation in Image Sequences
This paper proposes a method for depth estimation in video sequences acquired by a monocular came... more This paper proposes a method for depth estimation in video sequences acquired by a monocular camera mounted on a mobile platform. The proposed algorithm is able to estimate in real time the relative distances of the objects in the field of view exploiting the parallax effect, provided the platform movement complies with a few constraints. The developed system is designed to operate at the input pixel cadence and is thus applicable to any video resolution. The final architecture, using operators no more complex than an adder and a memory that is just a fraction of a frame memory, can be realized in a low-cost FPGA.
AtmoCube: observation of the near Earth space environment to study "space weather" effects
ABSTRACT AtmoCube represents an innovative measurement system for the study of the near Earth spa... more ABSTRACT AtmoCube represents an innovative measurement system for the study of the near Earth space environment starting from about 400 km altitude. It appears as a cubic nano-satellite, 13 cm side, with a total mass of 1.3 kg (aluminum structure). The scientific instrument is a spectral dosimeter monitoring the radiation environment. The payload also includes a magnetometer and a GPS. The goal is to build a precise map of the Earth magnetic field and of the flux of radiation impinging on the instrument, which is related to Space Weather effects. The GPS, providing the position of the region where the measurements are performed and the corresponding time instant, allows to correlate these measurements with other measurements performed on Earth and from the Space and therefore with the indexes of the Solar activity. In addition the GPS allows an indirect measurement of the atmospheric density.
Application-Specific Instruction-Set Processor for Retinex-Like Image and Video Processing
IEEE Transactions on Circuits and Systems II: Express Briefs, 2007
This brief presents an application-specific instruction-set processor (ASIP) for real-time Retine... more This brief presents an application-specific instruction-set processor (ASIP) for real-time Retinex image and video filtering. Design optimizations are addressed at algorithmic and architectural levels, the latter including a dedicated memory structure, an adapted pipeline, bypasses, a custom address generator and special looping structures. Synthesized in CMOS technology, the ASIP stands for its better energy-flexibility tradeoff versus reference ASIC and digital signal processing Retinex implementations.
Realization of an algorithm for image enhancement
A Face Recognition System Using Off-the-Shelf Feature Extractors and an Ad-Hoc Classifier
Face recognition systems are of great interest in many applications. We present some results from... more Face recognition systems are of great interest in many applications. We present some results from a comparison on different classification methods using an open source tool that works with Convolutional Neural Networks to extract facial features. This work focuses on the performance obtainable from a multi-class classifier, trained with a reduced number images, to identify a person between a group of known and unknown subjects . The overall system has been implemented in an Odroid XU-4 Platform.
A Novel Architecture for Real-Time Space Variant Image Deconvolution
Feeding a DNN for face verification in video data acquired by a visually impaired user
Some experiments on a face verification tool based on FaceNet are presented in this paper. The ta... more Some experiments on a face verification tool based on FaceNet are presented in this paper. The task of the system is to perform face verification in a real-time assistive system aiming at facilitating the approach between a blind person and a preselected acquaintance of his/her who enters the field of view. Face detection is made easier by the fact that an almost frontal view of the face is highly probable; verification on the contrary is difficult due to the poor quality of the acquired images and to the necessity of achieving a very low error rate. A custom database consisting of subjects required for verification is populated with face images provided by a suitable detection tool. The cascade of FaceNet and a Bayesian Classifier proves to be an effective tool for this unconstrained face verification task.
Computer vision for the blind: A dataset for experiments on face detection and recognition
We present a video dataset created for the needs of a research project that aims at developing vi... more We present a video dataset created for the needs of a research project that aims at developing vision-based techniques that assist the social interaction of a blind person. Two totally blind users have acquired the sequences, using at the same time a glasses-mounted camera and a necklace-mounted one. The video sequences were acquired in different environments, selecting conditions in which a user could be interested in detecting the presence of some of his/her acquaintances, to approach them in a most natural way. The sequences have been temporally cropped to extract video shots that, by inspection, were deemed to contain events valuable for the goals of the project. The shots are presently being annotated, and some preliminary experiments on face detection have been performed on the annotated data. We also present some observations about the specific application that is being considered.
This paper proposes a new neural network structure for image processing whose convolutional layer... more This paper proposes a new neural network structure for image processing whose convolutional layers, instead of using kernels with fixed coefficients, use space-variant coefficients. The adoption of this strategy allows the system to adapt its behavior according to the spatial characteristics of the input data. This type of layers performs, as we demonstrate, a non-linear transfer function. The features generated by these layers, compared to the ones generated by canonical CNN layers, are more complex and more suitable to fit to the local characteristics of the images. Networks composed by these non-linear layers offer performance comparable with or superior to the ones which use canonical Convolutional Networks, using fewer layers and a significantly lower number of features. Several applications of these newly conceived networks to classical image-processing problems are analyzed. In particular, we consider: Single-Image Super-Resolution (SISR), Edge-Preserving Smoothing (EPS), Noi...
Real-Time DNN-Based Face Identification for the Blind
Applications in Electronics Pervading Industry, Environment and Society, Jul 13, 2018
A New FPGA-Based Architecture for Iterative and Space-Variant Image Processing
Applications in Electronics Pervading Industry, Environment and Society, 2017
We propose a strategy to realize an innovative FPGA-based architecture able to speed up the Lucy-... more We propose a strategy to realize an innovative FPGA-based architecture able to speed up the Lucy-Richardson algorithm (LRA) for space-variant image deconvolution. The architecture exploits the possibility to distribute data into different memory blocks in the FPGA. In such a way, the algorithm execution is split into several channels operating in parallel. Since the LRA is implemented via an iterative and space-variant convolution, the approach adopted in this paper can be exploited in other similar image processing algorithms.
Face recognition functions are today exploited through biometric sensors in many applications, fr... more Face recognition functions are today exploited through biometric sensors in many applications, from extended security systems to inclusion devices; deep neural network methods are reaching in this field stunning performances. The main limitation of the deep learning approach is an inconvenient relation between the accuracy of the results and the needed computing power. When a personal device is employed, in particular, many algorithms require a cloud computing approach to achieve the expected performances; other algorithms adopt models that are simple by design. A third viable option consists of model (oracle) distillation. This is the most intriguing among the compression techniques since it permits to devise of the minimal structure that will enforce the same I/O relation as the original model. In this paper, a distillation technique is applied to a complex model, enabling the introduction of fast state-of-the-art recognition capabilities on a low-end hardware face recognition sen...
FPGA implementation of the Lucy-Richardson algorithm for fast space-variant image deconvolution
2015 9th International Symposium on Image and Signal Processing and Analysis (ISPA), 2015
The Lucy-Richardson algorithm is a very well-known method for non-blind image deconvolution. It c... more The Lucy-Richardson algorithm is a very well-known method for non-blind image deconvolution. It can also deal with space-variant problems, but it is seldom used in these cases because of its iterative nature and complexity of realization. In this paper we show that exploiting the sparse structure of the deconvolution matrix, and utilizing a specifically devised architecture, the restoration can be performed almost in real-time on VGA-size images.
The paper presents a novel technique for robust motion analysis in real automotive scenarios base... more The paper presents a novel technique for robust motion analysis in real automotive scenarios based on integrated Retinex-like pre-processing algorithm with block matching video motion estimator. Both algorithmic and real-time hardware design issues are discussed. The benefits of the proposed technique are manifold: the entire system is more robust; the estimated motion vectors are more reliable and less dependent on critical ambient conditions like shadows or flashes; the proposed algorithm may allow to perform motion estimation using very few bits and running as a 2-or 1-bit transform, still maintaining good performances. Real-time hardware implementation is achieved by design and synthesis in 65 nm CMOS standard-cells technology of an Application Specific Instruction-set Processor. Design optimizations for both the processing core and the memory organization are presented. With respect to the state of the art the proposed hardware implementation ensures bounded circuit complexity, low power consumption and reprogrammability of the technique.
Uploads
Papers by Stefano Marsi