Multichannel distance filter

Pappas, M.; Pitas, Ioannis

doi:10.1109/78.806087

Cited by 6 publications

(4 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this paper, we propose an application-driven vector ordering technique based on a spectral purity-based criterion, where each pixel vector in the hyperspectral image is ordered according to its spectral distance to other neighboring pixel vectors in the data. This type of ordering, which can be seen as a modification of the D-ordering available in the literature [ 18 ], has been found in previous work to be effective in capturing both spatial and spectral variability in hyperspectral data analysis [ 12 ]. An important ambiguity not sufficiently explored in previous work has to do with the fact that the ordering imposed above is not injective in general, i.e., two or more distinct vectors may output the same minimum or maximum distance.…”

Section: Multi-channel Mathematical Morphologymentioning

confidence: 99%

Multi-Channel Morphological Profiles for Classification of Hyperspectral Images Using Support Vector Machines

Plaza

Barra

2009

Sensors

View full text Add to dashboard Cite

Hyperspectral imaging is a new remote sensing technique that generates hundreds of images, corresponding to different wavelength channels, for the same area on the surface of the Earth. Supervised classification of hyperspectral image data sets is a challenging problem due to the limited availability of training samples (which are very difficult and costly to obtain in practice) and the extremely high dimensionality of the data. In this paper, we explore the use of multi-channel morphological profiles for feature extraction prior to classification of remotely sensed hyperspectral data sets using support vector machines (SVMs). In order to introduce multi-channel morphological transformations, which rely on ordering of pixel vectors in multidimensional space, several vector ordering strategies are investigated. A reduced implementation which builds the multi-channel morphological profile based on the first components resulting from a dimensional reduction transformation applied to the input data is also proposed. Our experimental results, conducted using three representative hyperspectral data sets collected by NASA's Airborne Visible-Infrared Imaging Spectrometer (AVIRIS) sensor and the German Digital Airborne Imaging Spectrometer (DAIS 7915), reveal that multi-channel morphological profiles can improve single-channel morphological profiles in the task of extracting relevant features for classification of hyperspectral data using small training sets.

show abstract

Section: Multi-channel Mathematical Morphologymentioning

confidence: 99%

Multi-Channel Morphological Profiles for Classification of Hyperspectral Images Using Support Vector Machines

Plaza

Barra

2009

Sensors

View full text Add to dashboard Cite

show abstract

“…The standard WD follows from (19) for the loss function F (e) = |e| 2 . The median form of the robust WD is obtained with F (e) = |e| [8], [18].…”

Section: B L-estimation Of the Wigner Distributionmentioning

confidence: 99%

“…The robust WD can be derived from the minimization problem (19), with the loss function F (e) = |e|. It assumes the form [7] W D I (n, k) = γ(n, k) 41is an implicit definition of W D I (n, k).…”

Section: Appendix Bmentioning

confidence: 99%

Robust L-estimation based forms of signal transforms and time-frequency representations

Djurović

Stanković

Böhme

2003

IEEE Trans. Signal Process.

View full text Add to dashboard Cite

“…To achieve ideal noise suppression and image detail preservation, weighted color vectors in stead of original ones are usually used in pixel ordering [46], [47] [48]. Adaptive vector filters, including adaptive nearest neighbor multichannel filters [49], fuzzy member functions based filters [50], multichannel distance filters [51], statistically computed neighborhood based filters [52], etc., in which the output is a weighted combination of all color vectors within the operation window, were also formulated. The key issue in this kind of nonlinear filters is optimal weight determination.…”

mentioning

confidence: 99%

Learning-based mathematical morphology in color image processing

Zhou¹

View full text Add to dashboard Cite

Mathematical Morphology is a powerful nonlinear image processing tool and is widely used in image segmentation, image filtering, shape representation, image measurement and so on. Most of the morphological operation based image processing approaches are developed for binary and grayscale images. Recently, since the color image has become the dominant image format nowadays and contains much more information than grayscale image, a problem of significant challenging in mathematical morphology is to extend basic morphological operations and their applications into color image processing area. This thesis mainly focuses on extending the existing basic morphological operators and their applications into color image processing area, by incorporating supervised and unsupervised learning techniques. The major contributions are summarized below: A new color pixel ordering scheme is established. Pixel ordering is the foundation of mathematical morphology. Color pixels are represented by multi-dimensional vectors, and the intensity-based pixel ordering scheme used in grayscale morphology is no longer applicable to vector ordering in color morphology. Therefore, a probabilistic framework for color pixel ordering is established, where the probability estimator is learned through supervised learning. Based on the probability-based color pixel order-ATTENTION: The Singapore Copyright Act applies to the use of this document. Nanyang Technological University Library in ing scheme, basic morphological operations: erosion, dilation, opening and closing are built. Two edge detectors are developed by using of such basic color morphological operations. One simple edge detector for color images containing simple colored shapes, uses the difference between elementary dilated and eroded color images as the edge detection result, which is similar as the edge detector for binary image. Color morphological gradient vector (CMGV) is introduced for image segmentation problems in complicated color images. Supervised learning techniques are used to extract boundary information from CMGV, and it is effective when separating specific objects from the images containing other objects rather than the one of interests. The multichannel filtering approach established on unsupervised learning-based color morphological operations for impulsive noise removal is developed in this thesis. By using color pixel ordering scheme, which is learned from the pre-estimation of impulsive noise, contaminated pixels are ordered as maximum or minimum ones within the operation window, in color erosion or dilation respectively. This character ensures that only uncontaminated color pixels are distributed as supremum or infimum in morphological operations, hence noise pixels are suppressed. The color morphological operations and their application algorithms are tested on both synthetic and real color images. For segmentation application, color histological images are used. For image filtering application, scenery color images are used.

show abstract

Multichannel distance filter

Cited by 6 publications

References 16 publications

Multi-Channel Morphological Profiles for Classification of Hyperspectral Images Using Support Vector Machines

Multi-Channel Morphological Profiles for Classification of Hyperspectral Images Using Support Vector Machines

Robust L-estimation based forms of signal transforms and time-frequency representations

Learning-based mathematical morphology in color image processing

Contact Info

Product

Resources

About