Switching median and morphological filter for impulse noise removal from digital images
Introduction
In the process of recording or transmission, digital images are often contaminated by impulse noise arising from faulty sensors, channel transmission errors and timing errors in analog-to-digital conversion [1], [2]. To improve the quality of degraded digital images, it is of great significance to adopt effective approaches to remove impulse noise from digital images.
Various filters have been proposed for the removal of impulse noise. Among them, the median filter [3] and the morphological filter [4], [5], [6] have been well known for their superior performance in noise suppression and edge preservation in comparison with linear filters. However, the standard median filter and the morphological filter are implemented on the entire image and modify both noise pixels and noise-free ones, therefore resulting in damage to image details. To address this problem, numerous switching-based filters have been proposed by firstly utilizing a noise detector to identify the noise pixels and then removing them using the local statistics based filter. Examples include progressive switching median (PSM) filter [7], adaptive center weighted median (ACWM) filter [8], directional difference based switching median filter (DDSM) [9], Laplace detector-based switching median (LDSM) filter [10], pixel-wise MAD-based (PWMAD) filter [11], switching median filter with boundary discriminative noise detection (BDND) [12], opening closing sequence (OCS) filter [13], fast switching median (FSM) filter [14], efficient edge-preserving (EEP) filter [15], convolution noise detection-based switching median (CNDSM) filter [16] and noise adaptive fuzzy switching median (NAFSM) filter [17]. Although these filters usually perform better than the median filter and the morphological filter, they tend to remove many important features or retain numerous impulses in the filtered images at high noise ratios.
To effectively restore the images corrupted by impulse noise especially at high noise ratios, we propose a switching morphological and median (SMM) filter. In the proposed SMM filter, noise pixels are identified by the morphological noise detector based on such morphological operator as erosion, dilation, opening and closing. The detected noise pixels are restored by combining the improved median filter with the conditional morphological filter. The advantage of the proposed filter in noise detection and noise removal has been demonstrated by extensive comparisons with numerous well-known decision-based filters operating on standard test images.
Section snippets
Morphological noise detector
It has been well known that the pixels corrupted by impulse noise will display intensity extremes in their neighborhood. Accordingly, the erosion operator and the dilation operator are adopted to identify all the corrupted pixels in that the two operators correspond to finding the minimum and maximum of the pixel values within a specified neighborhood, respectively.
Let the input image f and the structuring element b denote two discrete-valued functions defined on a two-dimensional discrete
Morphological and median filter
The hybrid filter combining the conditional morphological filter with the improved median filter will be adopted to remove the detected impulses. For any noise pixel (i, j) with η(i, j) = 1 in the image, the conditional erosion and conditional dilation can be defined as:
Based on the above two operators, the conditional opening and conditional closing can
Experimental results
To demonstrate the effectiveness of the proposed SMM filter, comparisons about noise detection performance and restoration performance are made among the BDND filter, the OCS filter, the FSM filter, CNDSM filter, NAFSM filter and the SMM filter. The 512 × 512 gray-level images such as Bridge, Boat, Barbara and Pepper are chosen as the test images. These images are corrupted by the salt-pepper impulses with the equal probability. Simulations are carried out for a wide range of noise ratios varying
Conclusion
In this paper, we have proposed an efficient switching median and morphological filter using the morphological noise detector for the removal of impulse noise. The morphological detector can realize accurate noise detection at the various noise ratios. The integrating of the improved median filter with the conditional morphological filter is very effective for removing the detected impulses from the image while preserving image details. Simulation results demonstrate that the proposed filter
Acknowledgments
This work was supported by the Natural Science Foundation of Hubei Province of China (Grant No. Q20131706) and the National Natural Science Foundation of China (NSFC) (grant: 61350001).
References (17)
- et al.
Impulse noise removal utilizing second-order difference analysis
Signal Process.
(2007) - et al.
The analysis of morphological filters with multiple structuring elements
Comput. Vis. Graph. Image Process.
(1990) - et al.
A new impulse detection and filtering method for removal of wide range impulse noises
Pattern Recognit.
(2009) - et al.
Noise adaptive soft-switching median filter
IEEE Trans. Image Process.
(2001) Efficient removal of impulse noise from digital images
IEEE Trans. Consumer Electron.
(2006)- et al.
Weighted morphological filter
Electron. Lett.
(1998) - et al.
An edge preserving noise smoothing technique using multi-scale morphology
Signal Process.
(2002) - et al.
Progressive switching median filter for the removal of impulse noise from highly corrupted images
IEEE Trans. Circuits Syst. II: Analog Digit. Signal Process.
(1999)
Cited by (28)
Moving object area detection using normalized self adaptive optical flow
2016, OptikCitation Excerpt :However it fails to detect all the components of foreground and this result in bad target recognition. MF can be used to remove flow noise, but in some cases, it over-smoothes the edge features [12]. Rashidha et al. [13] proposed a modified weighted median filter method to stop over smoothing, which also reduces the computational time by completely smoothing the detected motion boundaries.
An Iterative Adaptive Fuzzy Filter Using Alpha Trimmed Mean Elimination of High Density Salt and Pepper Noise in Images
2022, AIP Conference ProceedingsAutomatic handwritten character recognition of Devanagari language: a hybrid training algorithm for neural network
2022, Evolutionary IntelligenceOptimized Convolutional Neural Network for Tamil Handwritten Character Recognition
2022, International Journal of Pattern Recognition and Artificial IntelligenceRestoration and enhancement of breast ultrasound images using extended complex diffusion based unsharp masking
2022, Proceedings of the Institution of Mechanical Engineers, Part H: Journal of Engineering in Medicine