Showing papers on "Median filter published in 1981"

A theoretical analysis of the properties of median filters

Abstract: Necessary and sufficient conditions for a signal to be invariant under a specific form of median filtering are derived. These conditions state that a signal must be locally monotone to pass through a median filter unchanged. It is proven that the form of successive median filtering of a signal (i.e., the filtered output is itself again filtered) eventually reduces the original signal to an invariant signal called a root signal. For a signal of length L samples, a maximum of \frac{1}{2}(L - 2) repeated filterings produces a root signal.

A Separable Median Filter for Image Noise Smoothing

Abstract: This paper investigates some properties of the separable filter resulting from successive applications of a one-dimensional median filter on the rows and columns of an image. Although the output of this separable filter is not identical to the corresponding nonseparable two-dimensional median filter with a square window, its performance in image noise smoothing is close. In particular, its effectiveness in smoothing noise and its behavior with edges are characterized and compared with those of the two-dimensional median filter. It is shown that the separable filter has a much simpler implementation in real-time hardware (at video rates, for example).

Some statistical properties of median filters

Abstract: Some statistical properties of median filters are analyzed. It is shown that median filters can remove impulsive plus Gaussian white noise better than Hanning filters when the amplitude of the impulses is large or the energy of the Gaussian noise is relatively low. It is also shown that, unlike linear filters, median filters preserve sharp changes in signals when the noise energy is not too high.

Techniques For Speckle Noise Removal

Abstract: In this paper, several techniques to reduce speckle noise (more generally signal independent multiplicative noise) in images are studied. The techniques include gray scale modification, frame averaging, low-pass filtering in the intensity and density domains, and application of the short space spectral subtraction image restoration technique in the density domain. Some discussions on the theoretical basis of the techniques studied are given and their performances are illustrated by way of examples.

On Second Moment Properties of Median Filtered Sequences of Independent Data

Abstract: In this paper we present the derivation of the bivariate distribution function for median filtered sequences of independent, arbitrary, second-order random variables. This result is then used to qualitatively analyze second moment properties of median filtered data. The results hint towards a low-pass characteristic of the median filters and a low dependency of the output spectrum on input alphabet size and distribution.

Finding Edges in Noisy Scenes

Abstract: Edge detection in the presence of noise is a well-known problem. This paper examines an applications-motivated approach for solving the problem using novel techniques and presents a method developed by the authors that performs well on a large class of targets. ROC curves are used to compare this method with other well-known edge detection operators, with favorable results. A theoretical argument is presented that favors LMMSE filtering over median filtering in extremely noisy scenes. Simulated results of the research are presented.

Acoustic noise suppression in the context of a perceptual model

Abstract: An acoustic noise suppression algorithm has been developed which suppresses noise from speech by first filtering it into a set of signals which approximate the loudness components perceived by the auditory system. These signals are generated by passing the input stimulus waveform through a filter bank with frequency bandwidths which approximate the ear's critical bandwidths. The noise on each signal is then suppressed using spectral subtraction techniques in a domain of simulated perception. This approach to noise suppression retains the intelligibility produced by spectral subtraction methods while eliminating the accompanying musical quality.

On low-roundoff noise and low-sensitivity digital filter structures

Abstract: This note presents a simple alternative form for the error-feedback digital filter structure which, with some minor modification of assumptions, can also represent the low-sensitivity and low-noise structures of Agarwal and Burrus. This form can thus explain why the Agarwal-Burrus structures have low-roundoff noise and can be shown to have low sensitivity. Finally, this form is generalized to define a new class of low-noise and low-sensitivity realization structures for digital filters.

Linear versus logarithmic spatial filtering in the removal of multiplicative noises.

Abstract: Linear and logarithmic coherent spatial-filtering techniques are compared in the separation of multiplicative noise from an image in a coherent-optical image-processing system It is demonstrated that both techniques are effective in the removal of simple grating-type high-frequency noises, but that linear filtering cannot remove noise of spatial frequencies lower than 5 lines/cm, whereas logarithmic filtering can

The DFP 9200 Digital Noise Reducer, A Real-Time High-Resolution Digital Video Processing System For X-Ray Fluoroscopy

Abstract: A dedicated digital processor is described capable of digitizing a high resolution video signal from a fluoroscopic TV camera into an 810 x 600 matrix in real time. For less demanding applications, a 512 x 512 matrix can be substituted. The sampling clock frequency is 15 Megahertz giving a Nyquist bandwidth limit of 7.5 MHz. A 7 MHz phase equalized eliptical filter at the input prevents aliasing and the production of false artifacts in the picture. Eleven bit digital processing follows an 8 bit analog to digital converter. Noise reduction is accomplished by a one frame recursive filter in which the filter coefficients are adjusted by a patented motion detector on a pixel by pixel basis to reduce motion smear. The lower perceived noise permits X-ray dose reduction of 2 to 8 times while retaining high quality pictures. A noise reduced spot picture can be frozen by a foot controlled switch permitting a further reduction of dosage and eliminating the need for a troublesome disc recorder. This noise reduced picture can also be used as a subtraction mask in an optional version of the equipment. A minimum of front panel operator controls for best human interface is accomplished by the use of a programmed read only memories to control all functions including noise reduction and frame storage.

Autoregressive models for noisy speech signals

Abstract: Linear prediction is a well extended technique for transmission, synthesis and recognition. However when the signal is corrupted by noise, the estimation of the auto-regressive model is known to be biaised. This paper is devoted to methods allowing a reduction of this bias. We will consider first a global method, in which the Yule Walker equations are modified to take into account the variance of an additive white noise. The problem becomes non-linear and is solved recursively. In a second approach, we will examine a time - recursive method based on Kalman filtering.

Single-Threshold Processing for Orthogonal Likelihood-Ratio Detection with Application to FSK Fiber-Optic Communications

Abstract: The conditions under which processing reduces to a single threshold comparison is derived for the binary orthogonal likelihoodratio, detection of a signal embedded in additive, independent noise. The range of the signal and noise is assumed to be discrete. Optimal processing of the observation vector, as Well as two commonly used ad hoc processing schemes, are discussed. It is shown, in all three cases, that single-threshold processing is optimal, provided a weak condition on the noise is satisfied, independent of the signal. The results are presented in the context of a PCM/FSK photocounting fiber-optic communication system in which the noise is considered to be of the Poisson or avalanche types. Although single-threshold processing is shown to be optimal in most cases, examples are presented where single-threshold processing is not optimal.

Use of linear prediction in detecting narrowband signals in colored noise

Abstract: Use of the LPF in cascade with an FFT is demonstrated as a suboptimum realization of the optimum filter for maximization of instantaneous signal-to-rms noise power. Computer generated colored noise is used to compare the suboptimum filter with a conventional matched filter (designed for white noise only) in terms of output SNR for cases in which the signal and noise spectra overlap. Significant improvements in SNR are seen with the LPF - FFT cascade. The LPF coefficients are also used to form a non-causal pre-filter which is used in cascade with an FFT to process the same data and the results compared to the suboptimum filter results. Improvements in SNR are obtained, but thresholding the FFT output is seen to be more complicated.

Image detection and enhancement

Abstract: An optimal filtering approach is taken to the problem of detecting and enhancing a photographic image immersed in background or noise. The minimum mean square error criterion is shown to be not relevant to the optical employment of the Wiener filter. It is shown theoretically and demonstrated by simulation that, even without information about the background, a knowledge of the target spectrum alone makes it possible to find and display such a target without distortion while eliminating all uncorrelated background, despite the fact that target and background spectra actually overlap, and virtually without regard to noise level.

Research on Image Enhancement Algorithms.

Abstract: : Six specific image enhancement algorithms are analyzed, simulated and evaluated for potential real-time application to imagery obtained from target acquisition systems. Three processing methods are recommended as operator controlled options for application to the imagery. They are median filtering, gradient processing and locally adaptive gain control. The objective for this effort was to identify and examine methods for enhanced acquisition and handoff for small targets in a complex and cluttered background. (Author)

On estimation of noise variance in the presence of a narrow-band signal

Abstract: The procedure for noise variance estimation in the presence of a narrow-band signal with a priori unknown parameters is discussed. The fast suppression of a narrow-band signal prior to measurement of a noise variance is accomplished with a low-order adaptive notch filter (ANF) based on a linear prediction. It is shown that a good estimation of noise variance can be obtained over the time-limited interval in spite of the presence of a strong narrow-band signal. Computer simulation results which demonstrate the proposed algorithm performance are included.

An adaptive digital filter as a sinusoidal noise canceller

Abstract: An adaptive filter is capable of varying its parameters adaptively in response to change in the characteristics of noise components. Thus it is effective for filtering out the nonstationary noise. This paper proposes an adaptive moving average filter (AMAF) for the extraction of desirable signal from the mixture of low-frequency signal, white noise and narrowband sinusoidal noise. The implementation of AMAF is simply an extension of a moving average filter, and therefore is easy to implement and yet very effective for sinusoidal noise removal. As a special case, when the desired signal is a sinusoidal waveform, the design characteristics such as the noise attenuation index and the distortion rate of the AMAF are evaluated via computer simulation. The simulated results match the results obtained through theoretical analysis.

Median Filter Noise Improvement of Digital Imagery

Abstract: : Designs for fire control automatic target recognition systems require poise-free or low-noise digital imagery of visible and infrared targets. Median filter noise cleaning of digital imagery was investigated to determine the extent to which different size median filter windows will improve the signal-to- noise ratio of a digital image with the least deterioration or degradation to the shape or position of the edges of a target. Examples of median filtering and adaptive window filtering (AWMF) are presented before and after noise cleaning, and an improvement over AWMF is proposed to remedy its display deficiencies. Recommendations are offered on how to proceed to obtain more meaningful results and how to extend the simulation to two dimensions before performing test on actual digital imagery data.

Optimal array processing with 3-D random arrays

Abstract: In this paper, we derive the detection performance of an optimally beamformed 3-D random array system, assuming known Gaussian signal, noise, and interference statistics, and compare it with the performance of a conventional beamformed system. The receiver operating characteristics (ROC) are straightforward, nonlinear functions of the particular array realization and the second-order signal, noise, and interference statistics. Thus, a determination of the average ROCs (over the array ensemble) appears intractable. However, by making a reasonable approximation, we are able to derive an analytical expression for the average ROC for both conventional and optimal processing.