Bilateral filtering smooths images while preserving edges, by means of a nonlinear combination of nearby image values. The method is noniterative, local, and simple. It combines gray levels or colors based on both their geometric closeness and their photometric similarity, and prefers near values to distant values in both domain and range. In contrast with filters that operate on the three bands of a color image separately, a bilateral filter can enforce the perceptual metric underlying the CIE-Lab color space, and smooth colors and preserve edges in a way that is tuned to human perception. Also, in contrast with standard filtering, bilateral filtering produces no phantom colors along edges in color images, and reduces phantom colors where they appear in the original image.

/pdf/bilateral-filtering-for-gray-and-color-images-jnho720ytd.pdf

Bilateral filtering for gray and color images

国際生活機能分類(International Classification of Functioning, Disability and Health: ICF)にみた福祉・保健・医療の専門職協働における連携に関する貢献と課題

This paper proposes an edge-directed interpolation algorithm for natural images. The basic idea is to first estimate local covariance coefficients from a low-resolution image and then use these covariance estimates to adapt the interpolation at a higher resolution based on the geometric duality between the low-resolution covariance and the high-resolution covariance. The edge-directed property of covariance-based adaptation attributes to its capability of tuning the interpolation coefficients to match an arbitrarily oriented step edge. A hybrid approach of switching between bilinear interpolation and covariance-based adaptive interpolation is proposed to reduce the overall computational complexity. Two important applications of the new interpolation algorithm are studied: resolution enhancement of grayscale images and reconstruction of color images from CCD samples. Simulation results demonstrate that our new interpolation algorithm substantially improves the subjective quality of the interpolated images over conventional linear interpolation.

New edge-directed interpolation

In this paper, we develop the nonsubsampled contourlet transform (NSCT) and study its applications. The construction proposed in this paper is based on a nonsubsampled pyramid structure and nonsubsampled directional filter banks. The result is a flexible multiscale, multidirection, and shift-invariant image decomposition that can be efficiently implemented via the a trous algorithm. At the core of the proposed scheme is the nonseparable two-channel nonsubsampled filter bank (NSFB). We exploit the less stringent design condition of the NSFB to design filters that lead to a NSCT with better frequency selectivity and regularity when compared to the contourlet transform. We propose a design framework based on the mapping approach, that allows for a fast implementation based on a lifting or ladder structure, and only uses one-dimensional filtering in some cases. In addition, our design ensures that the corresponding frame elements are regular, symmetric, and the frame is close to a tight one. We assess the performance of the NSCT in image denoising and enhancement applications. In both applications the NSCT compares favorably to other existing methods in the literature

/pdf/the-nonsubsampled-contourlet-transform-theory-design-and-3a4lsit6lx.pdf

The Nonsubsampled Contourlet Transform: Theory, Design, and Applications

Обнаружение транспортных средств на изображениях загородных шоссе на основе метода Single shot multibox Detector

This paper presents a new general framework for information fusion based on fuzzy methods. In the proposed architecture, feature extraction and feature fusion are performed by adopting a class of (possibly multidimensional) membership functions which take care of the possibility and admissibility of the feature itself. Test cases of one-dimensional and image data fusion are presented. >

Fuzzy methods for multisensor data fusion

ARationalUnsharpMaskingTechniqueGiovanniRamponiandAndreaPoleselDEEI, University of Triestevia A. Valerio 10,I-34127Trieste,Italye-mail:ramp oni@univ.trieste.itFinal version Nov.1997, accepted for publication in theJournal of Electronic ImagingAbstractThelinearUnsharpMaskingtechniqueusedinimagecontrast enhancementis mo di ed in this pap er,by intro-ducing acontrol term expressedasarational function ofthelo calinputdata.Inthisway,noiseampli cation isavoided and, at the same time, oersho ot e ectson sharpedges are limited. Exp erimental results supp ort the valid-ity of the metho d, even as a prepro cessor for interp olationsystems.Keywords:Image enhancement, Unsharp Masking, Ra-tional lters, Nonlinear op erators.I.IntroductionThe technique of Unsharp Masking (UM) was intro ducedin photography to improve the quality of pictures by mak-ing their details crisp er; it consisted in optically subtractinga blurred copy of an image from the image itself.Its digi-tal version, due to its simplicity and relative e ectiveness,has b ecome a to ol of widespread use in the image pro cess-ing community, describ ed in any textb o ok [1] and includedin many available softwarepackages (e.g.,[2]).Indigitalimage manipulation, it can b erealized as shown in Fig.1,by pro cessing the image with an highpass lter (usually aLaplacian), multiplying the result by a scaling factor, andadding it to the original data.Notwithstanding its p opularity, this technique su ers fromtwo drawbacks which can signi cantly reduceits b ene ts:noise sensitivity and excessiveoversho ot on sharp details.TheformerproblemcomesfromthefactthatUMmetho dassignsanemphasistothehighfrequencycom-p onents of the input, amplifying a part of the sp ectrum inwhich the SNR is usually low.On the opp osite, wide andabrupt luminance transitions in the input image can pro-duce oversho ot e ects; these are put into further evidenceby the human visual system through the Mach band e ect[1].Various mo di cations have b een intro duced in the basicUM technique, in particular to reduce the noise ampli ca-tion problem.A quite trivial approach consists in substi-tuting a bandpass lter for the highpass one in Fig.1. Thisof course reduces noise e ects, but also precludes e ectivedetail enhancement in most images. In more sophisticatedapproaches,nonlinearop erators(orderstatistics,p olyno-mial, logarithmic) are used to generate the correction signalwhichisaddedtotheimage; wewill citeafewof them,without attempting a thorough overview of the eld.Amo di edLaplacian lter,calledtheorderstatistics(OS)Laplacian,isprop osedin[3];itsoutputprop or-tional to the di erence b etween the average and the medianof the pixels in a window. The resulting OS-UM algorithmis evaluated for its p erformance on a convex/concae edgemo del and on white Gaussian noise input signal, showingits robustness and its enhancing characteristics.Alternatively, a p olynomial approach can b e used.In [4],the Laplacian lter is replaced by a simple op erator basedon a generalization of the so-called Teager's algorithm. Un-der reasonable hyp otheses, this op erator approximates theb ehaviour of a lo cal-mean-weighted highpass lter, havingreduced high-frequency gain in dark image areas.Accord-ingtoWeb er'slaw[1],thesensitivityofhumanvi-sualsystemishigherindarkareas;hencetheprop osed lter intro duces a p erceptually smaller noise ampli cation,without diminishing theedge{enhancing capability of theUM metho d.Another p olynomial metho d, the Cubic UM(CUM)technique,hasb eendevised[5]:its purp oseis toamplify onlylo calluminancechangesduetotrueimagedetails.ThisisachievedbymultiplyingtheoutputoftheLaplacian lterbyacontrolsignalobtainedfromquadratic edge sensor.Still in the p olynomial framework,a class of quadratic lters is de ned in [6], where the lterco ecients at a given p osition in the image are calculatedbytakingintoaccountthegreyleveldistributionofsurrounding pixels.A Gaussian or an exp onential functionis used to reduce the contribution of pixel values the lumi-nance of which is di erent from that of the center pixel.Finally, to acquirea b ettercontrol on therangeof theimage brightness,theUMmetho d canb ecoupledto ho-momorphic ltering [1]:in this case, the image is rst con-vertedto thelogarithmic domain, thenUMis p erformedand the output is exp onentiated and scaled [7].To try and cop e with b oth the drawbacks indicated ab ove,i.e.noiseampli cation and excessiveoversho ots,alinearadaptive op erator is prop osed in [8]. The LMS technique isused to change the value of the scaling factor (in Fig.1)at each lo cation of the image; to this purp ose, the pixel tob e pro cessedis lab elled as b elonging to a smo oth area orto a medium{ or high{contrast area.Go o d quality resultscan b eachieved,at theexp enseof a relatively high com-putational complexity. In this pap er we attempt to obtaina similar result with a simpler, non{adaptive metho d.ThebasicUMschemeisstillusedhere,butarationalfunc-1

Rational unsharp masking technique

Mini-drones are increasingly used in video surveillance. Their areal mobility and ability to carry video cameras provide new perspectives in visual surveillance which can impact privacy in ways that have not been considered in a typical surveillance scenario. To better understand and analyze them, we have created a publicly available video dataset of typical drone-based surveillance sequences in a car parking. Using the sequences from this dataset, we have assessed five privacy protection filters via a crowdsourcing evaluation. We asked crowdsourcing workers several privacy- and surveillance-related questions to determine the tradeoff between intelligibility of the scene and privacy, and we present conclusions of this evaluation in this paper.

/pdf/privacy-in-mini-drone-based-video-surveillance-2xcljga8tf.pdf

Privacy in mini-drone based video surveillance

Nonlinear fuzzy operators for image processing

The bi-impulse response of a 1-D or 2-D quadratic Volterra filter is introduced as a mathematical tool able to completely describe the nonlinear operator. The conditions that must be satisfied in order to obtain isotropic input/output relations are studied. The result is a formal framework that allows simple but effective operators (particularly for image enhancement and preprocessing) to be designed and a deeper insight into the properties of Volterra filters to be acquired. Examples of design and of the performance obtained by processing natural and synthetic images are presented. >

Giovanni Ramponi

Papers

Fuzzy methods for multisensor data fusion

Rational unsharp masking technique

Privacy in mini-drone based video surveillance

Nonlinear fuzzy operators for image processing

Bi-impulse response design of isotropic quadratic filters