Hierarchy-Based Salient Regions: A Region Detector Based on Hierarchies of Partitions

Abstract

This article introduces a novel region detector based on hierarchies of partitions, so-called Hierarchy-Based Salient Regions (HBSR). This approach enables to combine the clues given by a high quality contour detector with a custom salient region detection procedure. The evaluation of the proposed method HBSR with a standard feature detection assessment framework shows that HBSR outperforms the state-of-the-art methods, in average. These promising results may lead to improvements in many computer vision tasks.

Graph-based image gradients aggregated with random forests

“…Wellknown gradient operators based on kernel filters for local variation, such as Laplacian and Sobel, used to be the preferable operators [13,14]. However, learned edge maps, as in the Structured Edge Detection (SED) method [15], were proven to produce better results [8] and became more popular [16,17].…”

A Comparative Study of Certain Classifiers for Bharatanatyam Mudra Images' Classification using Hu-Moments

“…Recentadvancementsindigitalimageprocessinghaveledtodevelopmentsofvariousapplications suchashandgesturerecognition (Liu,Yin,&Zhang,2012;Wang,Chen,&Li,2016),signlanguage identification (Karishma&Jalal,2013;Raut,Mali,Thepade,&Sanas,2014;Zadghorban&Nahvi, 2016;Fagiani,Principi,Squartini,&Piazza,2015;Zaki&Shaheen,2011),bodyposturerecognition (Dantone,2013),andhumanactionrecognition (Kishoreetal.,2018) Theshapesofmudrasareimportantintheclassification.Manyshapedescriptorshavebeen proposedfordifferentapplications.Popularlyusedshapedescriptors,moments,findapplications invariousareaslikecharacterrecognition,facerecognition, handgesturerecognition, andfacial expressionrecognition (Devi&Saharia,2018;Fernando&Wijayanayake,2015;Premaratne,Yang, Zou, & Vial, 2013;Solís, Martinez, & Espinoza, 2016;Otiniano-Rodríguez, Cámara-Chávez, & Menotti, 2012;Hu, 1962). It is being attempted to deploy Hu-moments for the recognition and classificationofBharatanatyammudraimages.Hu-momentsrepresentsuitableshapefeaturesbecause theyarescale,translation,androtationinvariant.Theseveninvariantmomentsarefitforthedescription ofoverallshapeofthetarget.Hencemomentsareusedasfeaturesinthiswork.…”