DOCSLIB.ORG

Image Restoration Using Gaussian Scale Mixtures in the Wavelet Domain

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Image Restoration Using Gaussian Scale Mixtures in the Wavelet Domain IMAGE RESTORATION USING GAUSSIAN SCALE MIXTURES IN THE WAVELET DOMAIN Javier Portilla ∗ Eero Simoncelli y Visual Information Processing Group Center for Neural Science, and Dept. of Comp. Science and Artif. Intell. Courant Inst. of Mathematical Sciences Universidad de Granada New York University, NY 10003 [email protected] [email protected] We describe a statistical model for images decomposed 2. IMAGE MODELING in an overcomplete wavelet pyramid. Each neighborhood of pyramid coefficients is modeled as the product of a Gaus- Modeling the statistics of natural images is a difficult task, sian vector of known covariance, and an independent hid- because of the high dimensionality of the signal and the den positive scalar random variable. We propose an effi- higher-order statistical dependencies that are prevalent. Sim- cient Bayesian estimator for the pyramid coefficients of an plifying assumptions, such as homogeneity and locality are image degraded by linear distortion (e.g., blur) and additive usually applied. In the past decade it has become standard Gaussian noise. We demonstrate the quality of our results to begin by decomposing the image with a set of multi-scale in simulations over a wide range of blur and noise levels. band-pass oriented filters. This kind of representation is adapted to the approximate scale invariance of natural im- 1. INTRODUCTION ages, and it has been shown to decouple some high-order statistics. Another technique for simplifying the descrip- Natural images have distinct features that allow the human tion of high-order statistics is to use a low-order local model visual system to detect the presence of distortion, and to (e.g., Gaussian) whose parameters (e.g., variance) are gov- extract remaining information from the observation. Im- erned by a local hidden random variable. In this work we age restoration aims to construct an approximation sharing have combined these ideas into a Bayesian framework. the relevant features still present in the corrupted image, but with the artifacts suppressed. In order to distinguish the ar- tifacts from the signal, a good image model is essential. In 2.1. Image representation this work we use a Bayesian framework for image restora- tion, basing on a prior statistical model for natural images. As a preprocessing stage for image modeling, we use a fixed We assume Gaussian noise of known covariance has been multi-scale multi-orientation linear decomposition. It is now added after having convolved the image with a known blur- generally agreed that the use of overcomplete representa- ring kernel. Such model for the distortion provides a reason- tions is advantageous for restoration, in order to avoid alias- able approximation to real-world corruption sources, such ing artifacts that plague critically-sampled representations as de-focus or photon and electron noise in cameras. In such as orthogonal wavelets [9]. A widely followed ap- previous works, we have described a model for wavelet co- proach is to use orthogonal or biorthogonal basis functions, efficients using scale mixtures of Gaussians [1, 2]. Related but without decimating the subbands [e.g., 10]. However, models have been developed by several groups [3, 4, 5, 6]. after the critical sampling constraint has been removed, sig- We have applied this model to estimate images in the pres- nificant improvement comes from using representations with ence of independent additive Gaussian noise of known co- higher redundancy and orientation selectivity [9, 11, 12]. variance, following a suboptimal empirical Bayes strat- For the current paper, we have used a particular variant of an egy [2, 7]. Recently, we developed a direct Bayesian Least overcomplete tight frame representation known as a “steer- Square denoising procedure [8]. In this paper, we extend able pyramid” [9] (see [8] for details). In this paper we have this approach to include the full restoration problem. used 5 scales and 4 orientations for the decomposition, for a total number of 25 subbands. This number includes 4 5 ∗JP is supported by a ”Ramon´ y Cajal” grant (Science and Technology × Spanish Ministry) band-pass subbands plus 4 oriented high-pass subbands and yEPS is supported by NSF CAREER grant MIP-9796040. a non-oriented low-pass residual. 2.2. Gaussian scale mixtures in the wavelet domain prior, which for a Gaussian variable with a scale parameter pz yields [16]: For each coefficient in the pyramid representation, we con- 1 sider a neighborhood of coefficients, referring to the center pz(z) : (3) / z coefficient as the reference coefficient of the neighborhood. To avoid computational problems at the estimation stage, The neighborhood may include coefficients from other sub- we have set the prior to zero in the interval , where bands (i.e., corresponding to basis functions at nearby scales [0; zmin) is a small positive number.1 This solution is simple and and orientations), as well as from the same subband. For zmin efficient to implement. Surprisingly, it also performs better this work we have used a 3 3 neighborhood around the than maximum likelihood fitted non-parametric priors for reference coefficient, plus the×parent coefficient (same ori- the neighborhoods of each subband.2 entation and position, next coarser scale), whenever it ex- ists. Due to sampling differences at different scales, the coarser subband must be resampled at double rate in both 3. RESTORATION dimensions for obtaining the parent of each coefficient of the subband. We model the observed image as: We use a Gaussian scale mixture (GSM) to model the y(n; m) = x(n; m) h(n; m) + w(n; m); (4) coefficients within each local neighborhood with reference ∗ coefficients belonging to every given subband. A random where ”*” denotes convolution, x is the original image, h vector x is a Gaussian scale mixture [13] if it can be ex- is a known linear kernel and w is Gaussian noise of known pressed as the product of a zero-mean Gaussian vector u power spectral density (PSD) P (u; v). and an independent positive scalar random variable pz: w x = pzu: (1) 3.1. GSM with blur and noise The variable z is the multiplier. Vector x is an infinite mix- We follow the usual procedure for image restoration in the ture of Gaussian vectors, whose density is determined by the wavelet domain: (1) decompose the image y into pyramid subbands; (2) restore each subband, except for the low-pass covariance matrix Cuu of u and the mixing density pz(z): residual; and (3) invert the pyramid transform. Translating (4) into our local GSM model, a vector y corresponding to a px(x) = p(x z) pz(z) dz Z j neighborhood of N observed coefficients around a reference T −1 coefficient for each subband of the pyramid representation exp x (zCuu) x=2 − (2) = N=2 1=2 pz(z) dz; can be expressed as: Z (2 π) zCuu j j where N is the dimensionality of x and u (the size of the y = Hx + w = pzHu + w; (5) neighborhood). Without loss of generality one can assume where u and w are zero-mean Gaussian vectors, and H is an E z = 1, which implies C = C . uu xx N M matrix expressing the convolution of N coefficients f GSMg densities are symmetric and zero-mean, and they with× a kernel h of size N (N >> N, typically).3 The have leptokurtotic marginal densities (i.e., heavier tails than h h density of the observed neighborhood vector conditioned on a Gaussian). Another key property of the GSM model is that z is zero-mean Gaussian, with covariance C = zC 0 0 + the density of x is Gaussian when conditioned on z. Also, yjz u u C : the normalized vector x=pz is Gaussian. They also present ww interesting joint statistics: the variance of a vector element T −1 exp y (zCu0u0 + Cww) y=2 conditioned on a neighbor scales roughly linearly with the p(y z) = − ; (6) N j (2π) zCu0u0 + Cww square of the neighbor’s value. These marginal and joint j j statistics of GSM distributions are qualitatively similar to p where C 0 0 = HC HT is the N N covariance ma- those of neighbor coefficients responding to natural images u u uu trix of u0 = Hu. As the noise PSD ×P (u; v) is assumed in multi-scale and multi-orientation representation [14, 1]. w known, Cww is easily estimated by computing, for each subband, the sample covariance for the pyramid coefficients 2.3. Prior density for multiplier 9 1We have chosen zmin = 10− in our implementation. Results remain 17 5 Our model requires a specification of the density of the mul- almost unaffected with zmin 2 [10− ; 10− ]. tiplier, z. An alternative to estimate adaptive priors for the 2Note that, when estimating parameters under an overcomplete repre- hidden multiplier (as in [e.g., 5, 7]) is to use a noninfor- sentation, the set of least squares optimal solutions for the subbands is not least square optimal for the whole image. mative prior [15], which does not require the fitting of any 3The kernel h must be cropped in the frequency domain according to parameters to the observation. We have applied Jeffrey’s the spatial resolution of the pyramid subband. −T T −1 of the inverse Fourier transform of Pw(u; v) (a delta for where M = Cxx0 S Q, and v = Q S y. Finally, we the case of white noise). Given Cwwpand taking the expec- restrict the estimate to the reference coefficient: C C 0 0 tation of yjz over z, the filtered signal covariance u u N can be computed from the observation covariance matrix zmcnvn E xc y; z = ; (11) Cyy: Cyy = E z Cu0u0 + Cww.
Load more

Recommended publications
  • Moving Average Filters
    CHAPTER 15 Moving Average Filters The moving average is the most common filter in DSP, mainly because it is the easiest digital filter to understand and use. In spite of its simplicity, the moving average filter is optimal for a common task: reducing random noise while retaining a sharp step response. This makes it the premier filter for time domain encoded signals. However, the moving average is the worst filter for frequency domain encoded signals, with little ability to separate one band of frequencies from another. Relatives of the moving average filter include the Gaussian, Blackman, and multiple- pass moving average. These have slightly better performance in the frequency domain, at the expense of increased computation time. Implementation by Convolution As the name implies, the moving average filter operates by averaging a number of points from the input signal to produce each point in the output signal. In equation form, this is written: EQUATION 15-1 Equation of the moving average filter. In M &1 this equation, x[ ] is the input signal, y[ ] is ' 1 % y[i] j x [i j ] the output signal, and M is the number of M j'0 points used in the moving average. This equation only uses points on one side of the output sample being calculated. Where x[ ] is the input signal, y[ ] is the output signal, and M is the number of points in the average. For example, in a 5 point moving average filter, point 80 in the output signal is given by: x [80] % x [81] % x [82] % x [83] % x [84] y [80] ' 5 277 278 The Scientist and Engineer's Guide to Digital Signal Processing As an alternative, the group of points from the input signal can be chosen symmetrically around the output point: x[78] % x[79] % x[80] % x[81] % x[82] y[80] ' 5 This corresponds to changing the summation in Eq.
    [Show full text]
  • Comparative Analysis of Gaussian Filter with Wavelet Denoising for Various Noises Present in Images
    ISSN (Print) : 0974-6846 Indian Journal of Science and Technology, Vol 9(47), DOI: 10.17485/ijst/2016/v9i47/106843, December 2016 ISSN (Online) : 0974-5645 Comparative Analysis of Gaussian Filter with Wavelet Denoising for Various Noises Present in Images Amanjot Singh1,2* and Jagroop Singh3 1I.K.G. P.T.U., Jalandhar – 144603,Punjab, India; 2School of Electronics and Electrical Engineering, Lovely Professional University, Phagwara - 144411, Punjab, India; [email protected] 3Department of Electronics and Communication Engineering, DAVIET, Jalandhar – 144008, Punjab, India; [email protected] Abstract Objectives: This paper is providing a comparative performance analysis of wavelet denoising with Gaussian filter applied variouson images noises contaminated usually present with various in images. noises. Wavelet Gaussian transform filter is ais basic used filter to convert used in the image images processing. to wavelet Its response domain. isBased varying on with its kernel sizes that have also been shown in analysis. Wavelet based de-noisingMethods/Analysis: is also one of the way of removing thresholding operations in wavelet domain noise could be removed from images. In this paper, image quality matrices like PSNR andFindings MSE have been compared for the various types of noises in images for different denoising methods. Moreover, the behavior of different methods for image denoising have been graphically shown in paper with MATLAB based simulations. : In the end wavelet based de-noising methods has been compared with Gaussian basedKeywords: filter. The Denoising, paper provides Gaussian a review Filter, MSE,of filters PSNR, and SNR, their Thresholding, denoising analysis Wavelet under Transform different noise conditions. 1. Introduction is of bell shaped.
    [Show full text]
  • MPA15-16 a Baseband Pulse Shaping Filter for Gaussian Minimum Shift Keying
    A BASEBAND PULSE SHAPING FILTER FOR GAUSSIAN MINIMUM SHIFT KEYING 1 2 3 3 N. Krishnapura , S. Pavan , C. Mathiazhagan ,B.Ramamurthi 1 Department of Electrical Engineering, Columbia University, New York, NY 10027, USA 2 Texas Instruments, Edison, NJ 08837, USA 3 Department of Electrical Engineering, Indian Institute of Technology, Chennai, 600036, India Email: [email protected] measurement results. ABSTRACT A quadrature mo dulation scheme to realize the Gaussian pulse shaping is used in digital commu- same function as Fig. 1 can be derived. In this pap er, nication systems like DECT, GSM, WLAN to min- we consider only the scheme shown in Fig. 1. imize the out of band sp ectral energy. The base- band rectangular pulse stream is passed through a 2. GAUSSIAN FREQUENCY SHIFT lter with a Gaussian impulse resp onse b efore fre- KEYING GFSK quency mo dulating the carrier. Traditionally this The output of the system shown in Fig. 1 can b e describ ed is done by storing the values of the pulse shap e by in a ROM and converting it to an analog wave- Z t form with a DAC followed by a smo othing lter. g d 1 y t = cos 2f t +2k c f This pap er explores a fully analog implementation 1 of an integrated Gaussian pulse shap er, which can where f is the unmo dulated carrier frequency, k is the c f result in a reduced power consumption and chip mo dulating index k =0:25 for Gaussian Minimum Shift f area. Keying|GMSK[1] and g denotes the convolution of the rectangular bit stream bt with values in f1; 1g 1.
    [Show full text]
  • A Simplified Realization for the Gaussian Filter in Surface Metrology
    In X. International Colloquium on Surfaces, Chemnitz (Germany), Jan. 31 - Feb. 02, 2000, M. Dietzsch, H. Trumpold, eds. (Shaker Verlag GmbH, Aachen, 2000), p. 133. A Simplified Realization for the Gaussian Filter in Surface Metrology Y. B. Yuan(1), T.V. Vorburger(2), J. F. Song(2), T. B. Renegar(2) 1 Guest Researcher, NIST; Harbin Institute of Technology (HIT), Harbin, China, 150001; 2 National Institute of Standards and Technology (NIST), Gaithersburg, MD 20899 USA. Abstract A simplified realization for the Gaussian filter in surface metrology is presented in this paper. The sampling function sinu u is used for simplifying the Gaussian function. According to the central limit theorem, when n approaches infinity, the function (sinu u)n approaches the form of a Gaussian function. So designed, the Gaussian filter is easily realized with high accuracy, high efficiency and without phase distortion. The relationship between the Gaussian filtered mean line and the mid-point locus (or moving average) mean line is also discussed. Key Words: surface roughness, mean line, sampling function, Gaussian filter 1. Introduction The Gaussian filter has been recommended by ISO 11562-1996 and ASME B46-1995 standards for determining the mean line in surface metrology [1-2]. Its weighting function is given by 1 2 h(t) = e−π(t / αλc ) , (1) αλc where α = 0.4697 , t is the independent variable in the spatial domain, and λc is the cut-off wavelength of the filter (in the units of t). If we use x(t) to stand for the primary unfiltered profile, m(t) for the Gaussian filtered mean line, and r(t) for the roughness profile, then m(t) = x(t)∗ h(t) (2) and r(t) = x(t) − m(t), (3) where the * represents a convolution of two functions.
    [Show full text]
  • Chapter 3 FILTERS
    Chapter 3 FILTERS Most images are a®ected to some extent by noise, that is unexplained variation in data: disturbances in image intensity which are either uninterpretable or not of interest. Image analysis is often simpli¯ed if this noise can be ¯ltered out. In an analogous way ¯lters are used in chemistry to free liquids from suspended impurities by passing them through a layer of sand or charcoal. Engineers working in signal processing have extended the meaning of the term ¯lter to include operations which accentuate features of interest in data. Employing this broader de¯nition, image ¯lters may be used to emphasise edges | that is, boundaries between objects or parts of objects in images. Filters provide an aid to visual interpretation of images, and can also be used as a precursor to further digital processing, such as segmentation (chapter 4). Most of the methods considered in chapter 2 operated on each pixel separately. Filters change a pixel's value taking into account the values of neighbouring pixels too. They may either be applied directly to recorded images, such as those in chapter 1, or after transformation of pixel values as discussed in chapter 2. To take a simple example, Figs 3.1(b){(d) show the results of applying three ¯lters to the cashmere ¯bres image, which has been redisplayed in Fig 3.1(a). ² Fig 3.1(b) is a display of the output from a 5 £ 5 moving average ¯lter. Each pixel has been replaced by the average of pixel values in a 5 £ 5 square, or window centred on that pixel.
    [Show full text]
  • Electronic Filters Design Tutorial - 3
    Electronic filters design tutorial - 3 High pass, low pass and notch passive filters In the first and second part of this tutorial we visited the band pass filters, with lumped and distributed elements. In this third part we will discuss about low-pass, high-pass and notch filters. The approach will be without mathematics, the goal will be to introduce readers to a physical knowledge of filters. People interested in a mathematical analysis will find in the appendix some books on the topic. Fig.2 ∗ The constant K low-pass filter: it was invented in 1922 by George Campbell and the Running the simulation we can see the response meaning of constant K is the expression: of the filter in fig.3 2 ZL* ZC = K = R ZL and ZC are the impedances of inductors and capacitors in the filter, while R is the terminating impedance. A look at fig.1 will be clarifying. Fig.3 It is clear that the sharpness of the response increase as the order of the filter increase. The ripple near the edge of the cutoff moves from Fig 1 monotonic in 3 rd order to ringing of about 1.7 dB for the 9 th order. This due to the mismatch of the The two filter configurations, at T and π are various sections that are connected to a 50 Ω displayed, all the reactance are 50 Ω and the impedance at the edges of the filter and filter cells are all equal. In practice the two series connected to reactive impedances between cells.
    [Show full text]
  • Memristor-Based Approximation of Gaussian Filter
    Memristor-based Approximation of Gaussian Filter Alex Pappachen James, Aidyn Zhambyl, Anju Nandakumar Electrical and Computer Engineering department Nazarbayev University School of Engineering Astana, Kazakhstan Email: [email protected], [email protected], [email protected] Abstract—Gaussian filter is a filter with impulse response of fourth section some mathematical analysis will be provided to Gaussian function. These filters are useful in image processing obtained results and finally the paper will be concluded by of 2D signals, as it removes unnecessary noise. Also, they could stating the result of the done work. be helpful for data transmission (e.g. GMSK modulation). In practice, the Gaussian filters could be approximately designed II. GAUSSIAN FILTER AND MEMRISTOR by several methods. One of these methods are to construct Gaussian-like filter with the help of memristors and RLC circuits. A. Gaussian Filter Therefore, the objective of this project is to find and design Gaussian filter is such filter which convolves the input signal appropriate model of Gaussian-like filter, by using mentioned devices. Finally, one possible model of Gaussian-like filter based with the impulse response of Gaussian function. In some on memristor designed and analysed in this paper. sources this process is also known as the Weierstrass transform [2]. The Gaussian function is given as in equation 1, where µ Index Terms—Gaussian filters, analog filter, memristor, func- is the time shift and σ is the scale. For input signal of x(t) tion approximation the output is the convolution of x(t) with Gaussian, as shown in equation 2.
    [Show full text]
  • Sampling and Reconstruction
    Sampling and reconstruction COMP 575/COMP 770 Spring 2016 • 1 Sampled representations • How to store and compute with continuous functions? • Common scheme for representation: samples write down the function’s values at many points [FvDFH fig.14.14b / Wolberg] [FvDFH fig.14.14b / • 2 Reconstruction • Making samples back into a continuous function for output (need realizable method) for analysis or processing (need mathematical method) amounts to “guessing” what the function did in between [FvDFH fig.14.14b / Wolberg] [FvDFH fig.14.14b / • 3 Filtering • Processing done on a function can be executed in continuous form (e.g. analog circuit) but can also be executed using sampled representation • Simple example: smoothing by averaging • 4 Roots of sampling • Nyquist 1928; Shannon 1949 famous results in information theory • 1940s: first practical uses in telecommunications • 1960s: first digital audio systems • 1970s: commercialization of digital audio • 1982: introduction of the Compact Disc the first high-profile consumer application • This is why all the terminology has a communications or audio “flavor” early applications are 1D; for us 2D (images) is important • 5 Sampling in digital audio • Recording: sound to analog to samples to disc • Playback: disc to samples to analog to sound again how can we be sure we are filling in the gaps correctly? • 6 Undersampling • What if we “missed” things between the samples? • Simple example: undersampling a sine wave unsurprising result: information is lost surprising result: indistinguishable from
    [Show full text]
  • Deconvolution
    Pulses and Parameters in the Time and Frequency Domain David A Humphreys National Physical Laboratory ([email protected]) Signal Processing Seminar 30 November 2005 Structure • Introduction • Waveforms and parameters • Instruments • Waveform metrology and correction Waveforms and parameters Parameters for a pulse waveform Signal Time Why use parameters? • Convenient way to compare information • Reduce data complexity • Specification of instruments • IEEE Standard on Transitions, Pulses, and Related Waveforms, IEEE Std 181, 2003 • These documents standardise how to characterise a waveform but do not specify units Measuring and characterising a signal Determine parameters that describe a signal or response I. Instrument capabilities II. Waveform metrology and correction III. Parameter analysis and extraction Assumptions • In practice we want to define a response in terms of single parameters such as width or risetime or bandwidth • Assume oscilloscope response has smooth bell-shaped e.g. Gaussian Simple analytically Can calculate frequency response, convolution simply, reasonably realistic Only 3 parameters – width/risetime; amplitude and position Instruments Instrumentation • Time-domain Frequency Real-time oscilloscope Time High-speed A/D converter Memory subsystem e.g. 32 Msamples X(t) A/D Memory Instrumentation • Time-domain Frequency Real-time oscilloscope Time Sample rates up to 40Gsamples/s Multiple high-speed A/D converters X(t) A/D Memory Fast memory subsystem A/D Memory Memory architecture may limit the length of trace A/D Memory e.g. 1 Msamples A/D Memory Resolution typically 8 bits Instrumentation • Time-domain Frequency Real-time oscilloscope Time Sampling oscilloscope Requires a repetitive waveform Sampling X(t) High-speed >70 GHz Gate Short trace length (e.g.
    [Show full text]
  • 3D Profile Filter Algorithm Based on Parallel Generalized B-Spline Approximating Gaussian
    CHINESE JOURNAL OF MECHANICAL ENGINEERING ·148· Vol. 28,aNo. 1,a2015 DOI: 10.3901/CJME.2014.1106.163, available online at www.springerlink.com; www.cjmenet.com; www.cjmenet.com.cn 3D Profile Filter Algorithm Based on Parallel Generalized B-spline Approximating Gaussian REN Zhiying1, 2, GAO Chenghui1, 2, *, and SHEN Ding3 1 School of Mechanical Engineering and Automation, Fuzhou University, Fuzhou 350108, China 2 Tribology Research Institute, Fuzhou University, Fuzhou 350108, China 3 Fujian Institute of Metrology, Fuzhou 35000, China Received January 24, 2014; revised September 10, 2014; accepted November 6, 2014 Abstract: Currently, the approximation methods of the Gaussian filter by some other spline filters have been developed. However, these methods are only suitable for the study of one-dimensional filtering, when these methods are used for three-dimensional filtering, it is found that a rounding error and quantization error would be passed to the next in every part. In this paper, a new and high-precision implementation approach for Gaussian filter is described, which is suitable for three-dimensional reference filtering. Based on the theory of generalized B-spline function and the variational principle, the transmission characteristics of a digital filter can be changed through the sensitivity of the parameters (t1, t2), and which can also reduce the rounding error and quantization error by the filter in a parallel form instead of the cascade form. Finally, the approximation filter of Gaussian filter is obtained. In order to verify
    [Show full text]
  • Multi-Focus Image Fusion Using Modified Gaussian Filter and Discrete Lifting Wavelet Transform Tarlok Singh1, Prof
    International Journal Of Electronics Engineering (Issn: 0973-7383) Volume 10 • Issue 2 pp. 201-209 June 2018-Dec 2018 www.csjournals.com Multi-focus Image Fusion using Modified Gaussian filter and Discrete Lifting Wavelet Transform Tarlok Singh1, Prof. Pammy Manchanda2 1, 2Guru Nanak Dev University, Amritsar, Punjab, India, 143001. [email protected] Abstract: Image fusion has always been a popular research area towards improving the visual quality of digital pictures. Selective picking of best parts from each image and integrating them into one work of art is called image fusion. We proposed a multi-focus image fusion method where partially blurred images are merged together and yield a better-quality image. This is done by applying fusion rules on obtained transformed images using discrete Lifting wavelet Transform (LWT). Proposed technique provides an agile ability for image fusion compared to the existing fusion techniques. The proposed algorithm is very simple to implement and faster than conventional techniques. Testing is done on a collection of standard images, with and without employing consistency verification. Keywords: Image fusion, Digital image processing, Lifting wavelet transform, Fusion rules, Visual sensor networks. I. Introduction Image fusion is the technique to blend more images while obtaining maximum information from the image and avoid noise or any fictional data [1]. The goal is to get one processed image of best quality compared to each individual input image. There are different categories of fusion such as Multi-focus, multi-sensor, single sensor, multi-view, multi-modal and multi-temporal [2]. In multi-focus image fusion method, multiple images of a given scene are obtained with focus on different objects.
    [Show full text]
  • Application Note AN-7A Copyright March, 1999
    Application Note AN-7a Copyright March, 1999 Low-Pass Risetime Filters for Time Domain Applications James R Andrews, PhD, IEEE Fellow When engineers start working in the Time Domain, they soon The single most important time domain instrument is the learn that the optimum pulse response is the Gaussian (see oscilloscope All oscilloscope manufacturers strive to make Figure 1) With the Gaussian waveshape, pulse signals rise their oscilloscopes with responses approximating the and fall smoothly and then rapidly settle to their final values Gaussian [1] With the Gaussian, their oscilloscopes come The presence of other perturbations on a pulse signal, such the closest to giving a true measure of the pulse signals as precursors, overshoot, and ringing are undesirable These under test with the least amount of extra artifacts introduced perturbations extend well beyond the risetime interval and by the scope When designing amplifiers and other create uncertainty in the actual value of the pulse for a much components for pulse applications, companies such as PSPL longer time duration With the Gaussian, one knows the final also strive to achieve Gaussian responses When engineers value very rapidly after the risetime The frequency response design digital telecommunication systems, they try to obtain of a Gaussian system is shown in Figure 2 For a pure Gaussian responses for all the various system components Gaussian, the time-frequency relationship is Tr*BW = 0332, Maintaining Gaussian responses helps to reduce inter- where Tr is the 10%-90% risetime,
    [Show full text]