This work addresses the problem of non-blind image deblurring for arbitrary input noise. The problem arises in the context of sensors with strong chromatic aberrations, as well as in standard cameras, in low-light and high-speed scenarios. A short description of two common classical approaches to regularized image deconvolution is provided, and common issues arising in this context are described. It is shown how a pre-deconvolved deep neural network (DNN) based image enhancement can be improved by joint optimization of regularization parameters and network weights. Furthermore, a two-step approach to deblurring based on two DNNs is proposed, with the first network estimating deconvolution regularization parameters, and the second one performing image enhancement and residual artifact removal. For the first network, a novel RegParamNet architecture is introduced and its performance is examined for both direct and indirect regularization parameter estimation. The system is shown to operate well for input noise in a three orders of magnitude range (0.01–10.0) and a wide spectrum of 1D or 2D Gaussian blur kernels, well outside the scope of most previously explored image blur and noise degrees. The proposed method is found to significantly outperform several leading state-of-the-art approaches.

Mapping vehicle surroundings using an occupancy grid is commonly a performance bottleneck in many automotive applications. In particular, three dimensional map generation becomes extremely complex because of the additional degree of freedom that significantly increases the amount of occupancy grid cells to be examined. In this work, we present a novel way to speed up generation of volumetric occupancy grids, by approximating heavy calculations and utilizing efficient data structures. Our proposed method produces a IS-fold speedup in a high-resolution mode and demonstrates realtime performance, without discernible impact on the accuracy. A description of the algorithm and the data structures is provided. The algorithm is tested on a multi-sensor system in a challenging environment. Moreover, our design is portable and could suit many platforms, including embedded automotive hardware.

Occupancy grids (OG) are widely used for low-level fusion of radar data in various automotive applications. At the core of OG generation, usually, there is an inverse sensor model (ISM), which is a conditional cell occupancy probability model. Traditional ISM's lack mechanisms decreasing occupancy likelihoods along directions that produce no detections; thus, false detections tend to perpetuate on the OG. In this paper, we propose a novel Inverse Sensor Model including a “positive” component describing occupancy probabilities induced by radar detections and a “negative” component handling lack of detections in a given direction. This dual model proves especially useful in multi-sensor/multi-frame context since false detections by different radars and/or at different moments are uncorrelated and thus can be efficiently mitigated.

Neuronal circuits' ability to maintain the delicate balance between stability and flexibility in changing environments is critical for normal neuronal functioning. However, to what extent individual neurons and neuronal populations maintain internal firing properties remains largely unknown. In this study, we show that distributions of spontaneous population firing rates and synchrony are subject to accurate homeostatic control following increase of synaptic inhibition in cultured hippocampal networks. Reduction in firing rate triggered synaptic and intrinsic adaptive responses operating as global homeostatic mechanisms to maintain firing macro-stability, without achieving local homeostasis at the single-neuron level. Adaptive mechanisms, while stabilizing population firing properties, reduced short-term facilitation essential for synaptic discrimination of input patterns. Thus, invariant ongoing population dynamics emerge from intrinsically unstable activity patterns of individual neurons and synapses. The observed differences in the precision of homeostatic control at different spatial scales challenge cell-autonomous theory of network homeostasis and suggest the existence of network-wide regulation rules.

A number of vital biological processes rely on fast and precise recognition of a specific DNA sequence (site) by a protein. How can a protein find its site on a long DNA molecule among 106–109 decoy sites? Here, we present our recent studies of the protein–DNA search problem. Seminal biophysical works suggested that the protein–DNA search is facilitated by 1D diffusion of the protein along DNA (sliding). We present a simple framework to calculate the mean search time and focus on several new aspects of the process such as the roles of DNA sequence and protein conformational flexibility. We demonstrate that coupling of DNA recognition with conformational transition within the protein–DNA complex is essential for fast search. To approach the complexity of the in vivo environment, we examine how the search can proceed at realistic DNA concentrations and binding constants. We propose a new mechanism for local distance-dependent search that is likely essential in bacteria. Simulations of the search on tightly packed DNA and crowded DNA demonstrate that our theoretical framework can be extended to correctly predicts search time in such complicated environments. We relate our findings to a broad range of experiments and summarize the results of our recent single-molecule studies of a eukaryotic protein (p53) sliding along DNA.

We consider self-avoiding polymers attached to the tip of an impenetrable probe. The scaling exponents gamma(1) and gamma(2), characterizing the number of configurations for the attachment of the polymer by one end, or at its midpoint, vary continuously with the tip's angle. These apex exponents are calculated analytically by epsilon expansion, and numerically by simulations in three dimensions. We find that when the polymer can move through the attachment point, it typically slides to one end; the apex exponents quantify the entropic barrier to threading the eye of the probe.

Many important transport phenomena are described by simple mathematical models rooted in the diffusion equation. Geometrical constraints present in such phenomena often have a global influence and manifest themselves in scaling relations and stable distribution functions. I treat a random walk confined to a half-space using several approaches: diffusion equations, lattice walks, and path integrals. Potential generalizations are discussed.

Recognition and binding of specific sites on DNA by proteins is central for many cellular functions such as transcription, replication, and recombination. In the process of recognition, a protein rapidly searches for its specific site on a long DNA molecule and then strongly binds this site. Here we aim to find a mechanism that can provide both a fast search (1-10 s) and high stability of the specific protein-DNA complex (Kd=10(-15)-10(-8) M). Earlier studies have suggested that rapid search involves sliding of the protein along the DNA. Here we consider sliding as a one-dimensional diffusion in a sequence-dependent rough energy landscape. We demonstrate that, despite the landscape's roughness, rapid search can be achieved if one-dimensional sliding is accompanied by three-dimensional diffusion. We estimate the range of the specific and nonspecific DNA-binding energy required for rapid search and suggest experiments that can test our mechanism. We show that optimal search requires a protein to spend half of its time sliding along the DNA and the other half diffusing in three dimensions. We also establish that, paradoxically, realistic energy functions cannot provide both rapid search and strong binding of a rigid protein. To reconcile these two fundamental requirements we propose a search-and-fold mechanism that involves the coupling of protein binding and partial protein folding. The proposed mechanism has several important biological implications for search in the presence of other proteins and nucleosomes, simultaneous search by several proteins, etc. The proposed mechanism also provides a new framework for interpretation of experimental and structural data on protein-DNA interactions.

Received 1 September 2004DOI:https://doi.org/10.1103/PhysRevE.70.049901©2004 American Physical Society

Many biological processes involve one-dimensional diffusion over a correlated inhomogeneous energy landscape with a correlation length xi(c). Typical examples are specific protein target location on DNA, nucleosome repositioning, or DNA translocation through a nanopore, in all cases with xi(c) approximately 10 nm. We investigate such transport processes by the mean first passage time (MFPT) formalism, and find diffusion times which exhibit strong sample to sample fluctuations. For a displacement N, the average MFPT is diffusive, while its standard deviation over the ensemble of energy profiles scales as N(3/2) with a large prefactor. Fluctuations are thus dominant for displacements smaller than a characteristic N(c) >> xi(c) : typical values are much less than the mean, and governed by an anomalous diffusion rule. Potential biological consequences of such random walks, composed of rapid scans in the vicinity of favorable energy valleys and occasional jumps to further valleys, is discussed.