scispace - formally typeset
Search or ask a question
Topic

Median filter

About: Median filter is a research topic. Over the lifetime, 12479 publications have been published within this topic receiving 178253 citations.


Papers
More filters
Journal ArticleDOI
TL;DR: In this paper, a novel regularization technique which can combine signals from all Global Positioning System (GPS) satellites for a given instant and a given receiver is developed to estimate the vertical total electron content (VTEC) values for the 24-hour period without missing any important features in the temporal domain.
Abstract: [1] A novel regularization technique which can combine signals from all Global Positioning System (GPS) satellites for a given instant and a given receiver is developed to estimate the vertical total electron content (VTEC) values for the 24-hour period without missing any important features in the temporal domain. The algorithm is based on the minimization of a cost function which also includes a high pass penalty filter. Optional weighting function and sliding window median filter are added to enrich the processing and smoothing of the data. The developed regularized estimation algorithm is applied to GPS data for various locations for the solar maximum week of 23–28 April 2001. The parameter set that is required by the estimation algorithm is chosen optimally using appropriate error functions. This robust and optimum parameter set can be used for all latitudes and for both quiet and disturbed days. It is observed that the estimated TEC values are in general accordance with the TEC estimates from other global ionospheric maps, especially for quiet days and midlatitudes. Owing to its 30 s time resolution, the regularized VTEC estimates from the developed algorithm are very successful in representation and tracking of sudden temporal variations of the ionosphere, especially for high latitudes and during ionospheric disturbances.

160 citations

Proceedings Article
24 May 2019
TL;DR: In this article, the authors propose a general framework for denoising high-dimensional measurements which requires no prior on the signal, no estimate of the noise, and no clean training data.
Abstract: We propose a general framework for denoising high-dimensional measurements which requires no prior on the signal, no estimate of the noise, and no clean training data. The only assumption is that the noise exhibits statistical independence across different dimensions of the measurement, while the true signal exhibits some correlation. For a broad class of functions ("$\mathcal{J}$-invariant"), it is then possible to estimate the performance of a denoiser from noisy data alone. This allows us to calibrate $\mathcal{J}$-invariant versions of any parameterised denoising algorithm, from the single hyperparameter of a median filter to the millions of weights of a deep neural network. We demonstrate this on natural image and microscopy data, where we exploit noise independence between pixels, and on single-cell gene expression data, where we exploit independence between detections of individual molecules. This framework generalizes recent work on training neural nets from noisy images and on cross-validation for matrix factorization.

158 citations

Proceedings ArticleDOI
16 Sep 1994
TL;DR: In this article, a decision rule based on the second order local statistics of the signal (within a window) is used to switch between the identity filter and a median filter, and the results on a test image show an improvement of around 4dB over the median filter alone, and 2dB over other techniques.
Abstract: Noise removal is important in many applications. When the noise has impulsive characteristics, linear techniquesdo not perform well, and median filter or its derivatives are often used. Although median-based filters preserve edgesreasonably well, they tend to remove some of the finer details in the image. Switching schemes — where the filter isswitched between two or more filters — have been proposed, but they usually lack a decision rule efficient enough toyield good results on different regions of the image. In this paper we present a strategy to overcome this problem. Adecision rule based on the second order local statistics of the signal (within a window) is used to switch between theidentity filter and a median filter. The results on a test image show an improvement of around 4dB over the medianfilter alone, and 2dB over other techniques.Keywords: Median filter; Image enhancement; Noise removal; Impulsive noise. 1. INTRODUCTION Noise reduction is often necessary as a pre-processing step in situations where a signal is contaminated by noise.In cases where the noise can be adequately modeled as additive Gaussian noise, linear filters are normally efficiciitfor noise-reduction. However, in many cases the noise is impulsive, and in this case linear techniques do not usuallyperform well. The median filter and its derivatives are often the filter of choice for these applications.The median filter is a non-linear filter, and it has the useful property of removing (reducing) impulsive noisewithout (severely) smoothing the edges of the signal. The main drawback of the median filter is that it also modifiesthe points not contaminated by noise, therefore removing the finer details in the signal.In the past 20 years, median filters have been generalized and modified in many ways. A good overview of pastwork on generalizations of median filters can be find in the paper by Gabbouj et al.1 Examples include rank orderfilters, weighted median filters, stack filters, and linear combinations of nonlinear filters. A theory for optimal stackfilters has been developed.2 More recently, filters where the rank selected is based on the pixel rank have been alsoproposed .

157 citations

Journal ArticleDOI
TL;DR: The distribution of the output of the one-dimensional median filter is derived for several cases including the k th-order output distribution with any input distribution, which is then used in several illustrative examples of median filtering a signal plus white noise.
Abstract: The distribution of the output of the one-dimensional median filter is derived for several cases including the k th-order output distribution with any input distribution. This is then used in several illustrative examples of median filtering a signal plus white noise.

157 citations

Journal ArticleDOI
07 Aug 2002
TL;DR: In this paper, two perceptual experiments that compare the perceptual quality of the output of different demosaicing algorithms are reported and it is found that a Bayesian demosaice algorithm produced the most preferred images.
Abstract: Demosaicing is an important part of the image-processing chain for many digital color cameras. The demosaicing operation converts a raw image acquired with a single sensor array, overlaid with a color filter array, into a full-color image. In this paper, we report the results of two perceptual experiments that compare the perceptual quality of the output of different demosaicing algorithms. In the first experiment, we found that a Bayesian demosaicing algorithm produced the most preferred images. Detailed examination of the data, however indicated that the good performance of this algorithm was at least in part due to the fact that it sharpened the images while it demosaiced them. In a second experiment, we silenced image sharpness as a factor by applying a sharpening algorithm to the output of each demosaicing algorithm. The optimal amount of sharpening to be applied to each image was chosen using the results of a preliminary experiment. Once sharpness was equated in this way, an algorithm developed by Freeman based on bilinear interpolation combined with median filtering, gave the best results. An analysis of our data suggests that our perceptual results cannot be easily predicted using an image metric.

156 citations


Network Information
Related Topics (5)
Feature extraction
111.8K papers, 2.1M citations
92% related
Image processing
229.9K papers, 3.5M citations
91% related
Convolutional neural network
74.7K papers, 2M citations
87% related
Artificial neural network
207K papers, 4.5M citations
86% related
Deep learning
79.8K papers, 2.1M citations
85% related
Performance
Metrics
No. of papers in the topic in previous years
YearPapers
202372
2022186
2021276
2020387
2019478
2018538