1,721,081 research outputs found
MACHINE LEARNING SOLUTIONS FOR OBJECTIVE VISUAL QUALITY ASSESSMENT
No full textObjective metrics for visual quality assessment usually improve their reliability by explicitly modeling the highly non-linear behavior of human perception; as a result, they often are complex, and computationally expensive. Conversely, Machine Learning (ML) paradigms allow to tackle the quality assessment task from a different perspective, as the eventual goal is to mimic quality perception instead of designing an explicit model the Human Visual System (HVS). Several studies already proved the ability of ML-based approach to address visual quality assessment. Indeed, a prerequisite for successfully using ML in modeling perceptual mechanisms is a profound understanding of the advantages and limitations that characterize learning machines. This paper illustrates and exemplifies the good practices to be followe
Hausdorff distance for robust and adaptive template selection in visual target detection
No full text
Embedded Electronic System Based on Dedicated Hardware DSPs for Electronic Skin Implementation
No full textAbstractThe effort to develop an electronic skin is highly motivated by many application domains namely robotics, biomedical instrumentations, and replacement prosthetic devices. Several e-skin systems have been proposed recently and have demonstrated the need of an embedded electronic system for tactile data processing either to mimic the human skin or to respond to the application demands. Processing tactile data requires efficient methods to extract meaningful information from raw sensors data.In this framework, our goal is the development of a dedicated embedded electronic system for electronic skin. The embedded electronic system has to acquire the tactile data, process and extract structured information. Machine Learning (ML) represents an effective method for data analysis in many domains: it has recently demonstrated its effectiveness in processing tactile sensors data.This paper presents an embedded electronic system based on dedicated hardware implementation for electronic skin systems. It provides a Tensorial kernel function implementation for machine learning based on Tensorial kernel approach. Results assess the time latency and the hardware complexity for real time functionality. The implementation results highlight the high amount of power consumption needed for the input touch modalities classification task. Conclusions and future perspectives are also presented
Enhanced Montgomery multiplication on DSP architectures for embedded public-key cryptosystems
No full textMontgomery's algorithm is a popular technique to speed up modular multiplications in public-key cryptosystems. This paper tackles the efficient support of modular exponentiation on inexpensive circuitry for embedded security services and proposes a variant of the finely integrated product scanning (FIPS) algorithm that is targeted to digital signal processors. The general approach improves on the basic FIPS formulation by removing potential inefficiencies and boosts the exploitation of computing resources. The reformulation of the basic FIPS structure results in a general approach that balances computational e fficiency and flexibility. Experimental results on commercial DSP platforms confirm both the method's validity and its effectiveness
Hybrid neural systems for reduced-reference image quality assessment
No full textReduced-reference paradigms are suitable for supporting real-time modeling of perceived quality, since they make use of salient features both from the target image and its original, undistorted version, without requiring the full original information. In this paper a reduced-reference system is proposed, based on a feature-based description of images which encodes relevant information on the changes in luminance distribution brought about by distortions. Such a numerical description feeds a double-layer hybrid neural system: first, the kind of distortion affecting the image is identified by a classifier relying on Support Vector Machines (SVMs); in a second step, the actual quality level of the distorted image is assessed by a dedicated predictor based on Circular Back Propagation (CBP) neural networks, specifically trained to assess image quality for a given artifact. The general validity of the approach is supported by experimental validations based on subjective quality data
- …
