An industrial portrait background removal solution based on knowledge infusion

Background removal of an identity (ID) picture consists in separating the foreground (face, body, hair and clothes) from the background of the image. It is a necessary groundwork for all modern identity documents that also has many benefits for improving ID security. State of the art image processing techniques encountered several segmentation issues and offer only partial solutions. It is due to the presence of erratic components like hairs, poor contrast, luminosity variation, shadow, color overlap between clothes and background. In this paper, a knowledge infused approach is proposed that hybridizes smart image processing tasks and prior knowledge. The research is based on a divide and conquer strategy aiming at simulating the sequential attention of human when performing a manual segmentation. Knowledge is infused by considering the spatial relation between anatomic elements of the ID image (face feature, forehead, body and hair) as well as their “signal properties”. The process consists in first determining a convex hull around the person’s body including all the foreground while keeping very close to the contour between the background and the foreground. Then, a body map generated from biometric analysis associated to an automatic grab cut process is applied to reach a finer segmentation. Finally, a heuristic-based post-processing step consisting in correcting potential hair and fine boundary issues leads to the final segmentation. Experimental results show that the newly proposed architecture achieves better performances than tested current state-of-the-art methodologies including active contours, generalist popular deep learning techniques, and also two other ones considered as the smartest for portrait segmentation. This new technology has been adopted by an international company as its industrial ID foreground solution.

     

Anisotropy measurements obtained by fractal analysis of trabecular bone at the calcaneus and radius

Antidepressant drugs have in common a delayed onset of clinical efficacy. In rats, long-term, daily administration of four different types of clinically effective antidepressant drugs results in decreased corticotropin releasing hormone (CRH) mRNA expression levels in the hypothalamic paraventricular nucleus (PVN). Because a subpopulation of neuropeptide Y (NPY) and proopiomelanocortin (POMC) neurons in the hypothalamic arcuate nucleus (Arc) projects to the PVN, we measured NPY and POMC mRNA expression in the Arc using in situ hybridization histochemistry at several time points following daily administration of four different antidepressant drugs. After 14 and 56 days of imipramine treatment, Arc NPY mRNA levels are decreased to 85% and 75% of control levels, but are unchanged compared to control after one or five days of treatment. Arc POMC mRNA levels are unchanged compared to controls at 1, 5, 14, or 56 days following imipramine treatment. Unlike after imipramine, Arc NPY and POMC mRNA levels are increased significantly to 134-172% of control following 56-day treatment with the antidepressant drugs fluoxetine, phenelzine, or idazoxan. The divergent effects of imipramine vs the other 3 antidepressant drugs on Arc NPY mRNA expression are similar to the pattern of changes in tyrosine hydroxylase (TH) mRNA expression levels in the locus coeruleus (LC) using the same experimental paradigm, but are different from the unidirectional depressive effects of all four drugs on CRH mRNA expression in the PVN. Thus, the Arc NPY and LC noradrenergic systems may act coordinately in mediating antidepressant effects. The present data are consistent with the delayed onset of clinical efficacy for antidepressant drugs, and suggest that Arc NPY and POMC neurotransmitter systems play a role in the pathophysiology of depression.     

Electrical behaviour of fractal nanosized tin dioxide films prepared by electrodeposition for gas sensing applications

An original electrochemical process to prepare SnO₂ gas sensors is detailed and correlated to electrical behaviour under gas environment. In particular conditions, Tin material was electrodeposited on insulating substrate to form a thin film principally composed of a single layer of individual nanoaggregates (5-10 nm in size). After tin electrodeposition, these supported aggregates were oxidised at air or pressurized oxygen to induce the formation of a fractal SnO₂ film. From these resulting active films, electrical measurements were carried out in ethanol and 300 ppm CO atmospheres. The results show, a sensitivity of 400% at 227 °C in the ethanol case with a response time of 140 s. When the temperature of electrical measurements increases, response and recovery times decrease. However, the sensing amplitude was not modified (Sensitivity around 4) between 250 and 300 °C. In the case of CO, the sensor presented a typical response with a factor of about 2.5 at 250 °C. A fractal dimension between 1.4 and 1.6 is found for fractal-shaped samples allowing an increase of specific surface in contact with gases. However, its does not effect sensitivity, which depends mainly on grain size.     

Biomechanical properties of human os calcanei: relationships with bone density and fractal evaluation of bone microarchitecture

The relationship between bone strength and bone mass is well established. The link between trabecular microarchitecture and biomechanical properties has been less extensively explored. To address this question, we have tested the mechanical behaviour of calcaneus bone samples and investigated the correlations between mechanical properties on the one hand, bone density and fractal analysis of microarchitecture on the other hand. Mechanical properties of 43 human os calcanei were determined by uniaxial compression testing of samples from tuber calcanei. Ash density, bulk density and dual energy X-ray absorptiometry of the samples were measured. Fractal analysis of the trabecular bone on calcaneus radiographs was performed by two estimators derived from the fractional Brownian motion model. The mechanical properties of human os calcis were found to correlate with age and density measurements. Fractal parameters derived from the bone texture analysis showed weaker but significant correlations with bone strength. Fractal analysis of texture could account in part for the variations of bone strength, but in this study cannot explain better than density the mechanical properties of trabecular bone. Nevertheless, it provides a non-invasive means of assessing molecular bone microarchitecture.     

An image dependent stopping method for iterative denoising procedures

Iterative methods are very successful for denoising images corrupted by random valued impulse noise. However, choosing the optimal number of iterations is a difficult issue. In this letter, a stopping method is proposed: the iterative denoising process is stopped when the number of cleaned pixels is minimal. It corresponds to the moment when the denoising process tends to modify noise-free pixels. It also corresponds with a high precision to the maximum of PSNR of the restored image. The originality of the method is that no a priori iteration number is to be chosen but the method results from image information. The proposed stopping strategy is therefore an efficient and image dependent method that can be easily implemented on real data.     

GPS/Galileo Multipath Mitigation Using the First Side Peak of Double Delta Correlator

In this paper, we propose an efficient scheme for the Line of Sight (LOS) delay estimation in satellite telecommunications. It aims to reduce errors caused by the Multipath (MP) fading channels in Global Navigation Satellite System (GNSS) such as Global Positioning System (GPS) and future Galileo. The scheme is based on the use of first side peaks of Double Delta Correlator (DDC) in combination with Savitzky-Golay Filter (SGF). The proposed scheme is called “Enhanced DDC (EDDC)”. The DDC technique, like High Resolution Correlator (HRC) and Strobe Correlator (SC), uses the central peak of the Correlation Function (CF) for the LOS delay estimation. As a result, it enables to deal with the medium-delay MP. Yet, it presents a greater sensitivity to the noise and to the long-delay MP that we propose to mitigate, respectively, with the use of SGF and the measure of the first side peak (instead of the central). The obtained results show that our proposed scheme gives better performances over DDC scheme. In fact, it shows a significant error reduction caused by the presence of long-delay MP signals (up to 90%) compared to that of the DDC scheme.     

A Microprocessor-Based Polarity Correlator for the On-Line Measurement of Muscle Fiber Conduction Velocity

The measurement of the average conduction velocity of action potentials along muscle fibers has important applications in muscle fatigue study and ergonomics. A cross-correlation-based method for the estimation of this velocity from surface electromyography (EMG) is presented. It is implemented on a standard low-cost 8-bit microprocessor. The use of cross correlation for such an on-line application has been made possible by the use of an efficient polarity correlation algorithm.