FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2017017507> ?p ?o ?g. }

• W2017017507 abstract "ABSTRACTA nonlinear correlation detection method for improving the selectivity in pattern recognition is presented. The new approach is based on morphological correlation. We propose a localized slicing carried on low and high component Fourierdomain informations.Keywords: Morphological Correlation, Pattern Recognition, Threshold Decomposition. 1. INTRODUCTION Correlation is one of the most popular approaches used in machine vision and pattern recognitio&. Nowadays, due toadvance in these applications, a high selectivity in recognition process is required. Matched filtering is already widely useof template matching and pattern recognition. However, the discrimination abilities of matched filters are not one of theirmore significant properties. Moreover, the common matched filter is optimum in the mean square error (MSE) criterion,meaning that minimizing the MSE leads to maximizing the common linear correlation (LC)2. The MSE is a function of L2norm and the acceptance of the common linear correlation is due to its Mathematical tractability of squared error metric andeasy optical implementation. In spite of these advantages, L' matching metric norm has been receiving lately wide use forits relevance in practical tasks. The mean absolute error (MAE) is one of the more representative metrics in L' norm space.It has been shown that minimising the MAE criterion is equivalent to maximising a nonlinear correlation namely, themorphological correlation (MC)3. This MC presents high discrimination abilities in pattern recognition as compared withthe common LC. Also, due to its nonlinearity property, MC offers a more robust detection of low-intensity images in thepresence of high intensity patterns to be rejected, moreover MC have been implemented optically4 taking advantages of theinherent parallelism of Optics.The MC can be expressed in term of a threshold decomposition concept5. It can be viewed as the sum over all amplitudes ofthe linear correlation between thresholded versions of an input scene and a reference pattern to be searched for in the inputscene, at every grey level3. In this paper we propose a localized threshold decomposition carried in the Fourier domain. Wetake profit of the direct correspondence to the frequency information of the images, and we apply the MC regarding withthat spectral analysis. The method relies on a partition in two recognition channels. The final combination of the correlationoutputs provides a high improvement in discrimination abilities with respect to the conventional MC. In all the simulationexperiments we deal with noise free images.In Section 2, we review the defmition of the MC based on threshold decomposition concept. Section 3 presents the localFourier domain slicing that we apply to the MC. Based on that, the novel nonlinear correlation technique for improving the" @default.
• W2017017507 created "2016-06-24" @default.
• W2017017507 creator A5010625698 @default.
• W2017017507 creator A5018345453 @default.
• W2017017507 creator A5057896714 @default.
• W2017017507 date "1998-05-22" @default.
• W2017017507 modified "2023-10-13" @default.
• W2017017507 title "Fourier localized threshold decomposition for nonlinear correlations" @default.
• W2017017507 doi "https://doi.org/10.1117/12.308926" @default.
• W2017017507 hasPublicationYear "1998" @default.
• W2017017507 type Work @default.
• W2017017507 sameAs 2017017507 @default.
• W2017017507 citedByCount "0" @default.
• W2017017507 crossrefType "proceedings-article" @default.
• W2017017507 hasAuthorship W2017017507A5010625698 @default.
• W2017017507 hasAuthorship W2017017507A5018345453 @default.
• W2017017507 hasAuthorship W2017017507A5057896714 @default.
• W2017017507 hasConcept C102519508 @default.
• W2017017507 hasConcept C121332964 @default.
• W2017017507 hasConcept C124681953 @default.
• W2017017507 hasConcept C134306372 @default.
• W2017017507 hasConcept C158622935 @default.
• W2017017507 hasConcept C178790620 @default.
• W2017017507 hasConcept C185592680 @default.
• W2017017507 hasConcept C203024314 @default.
• W2017017507 hasConcept C33923547 @default.
• W2017017507 hasConcept C41008148 @default.
• W2017017507 hasConcept C62520636 @default.
• W2017017507 hasConceptScore W2017017507C102519508 @default.
• W2017017507 hasConceptScore W2017017507C121332964 @default.
• W2017017507 hasConceptScore W2017017507C124681953 @default.
• W2017017507 hasConceptScore W2017017507C134306372 @default.
• W2017017507 hasConceptScore W2017017507C158622935 @default.
• W2017017507 hasConceptScore W2017017507C178790620 @default.
• W2017017507 hasConceptScore W2017017507C185592680 @default.
• W2017017507 hasConceptScore W2017017507C203024314 @default.
• W2017017507 hasConceptScore W2017017507C33923547 @default.
• W2017017507 hasConceptScore W2017017507C41008148 @default.
• W2017017507 hasConceptScore W2017017507C62520636 @default.
• W2017017507 hasLocation W20170175071 @default.
• W2017017507 hasOpenAccess W2017017507 @default.
• W2017017507 hasPrimaryLocation W20170175071 @default.
• W2017017507 hasRelatedWork W1980783579 @default.
• W2017017507 hasRelatedWork W1984031140 @default.
• W2017017507 hasRelatedWork W2034082304 @default.
• W2017017507 hasRelatedWork W2064118903 @default.
• W2017017507 hasRelatedWork W2068943040 @default.
• W2017017507 hasRelatedWork W2493392181 @default.
• W2017017507 hasRelatedWork W3199719804 @default.
• W2017017507 hasRelatedWork W3200098025 @default.
• W2017017507 hasRelatedWork W4231429051 @default.
• W2017017507 hasRelatedWork W81470468 @default.
• W2017017507 isParatext "false" @default.
• W2017017507 isRetracted "false" @default.
• W2017017507 magId "2017017507" @default.
• W2017017507 workType "article" @default.