Superpixel Segmentation by Object-Based Iterative Spanning Forest

Belém, Felipe; Guimarães, Silvio Jamil Ferzoli; Falcão, Alexandre X.

doi:10.1007/978-3-030-13469-3_39

Cited by 14 publications

(38 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…In this section, we present the Object-based ISF (OISF) [8] framework, which is a generalization of the Iterative Spanning Forest (ISF) [7] framework. The ISF is an efficient three-staged superpixel segmentation framework in which each component can be defined independently, being the major reason for recent publications shortly after its own [19]- [21].…”

Section: Object-based Superpixel Segmentationmentioning

confidence: 99%

“…In [8], we present the Object Geodesic Grid Sampling (OGRID) method in which samples seeds equidistantly within the probable objects in the map. The set of probable objects C is obtained by thresholding for a given minimum certainty value t. Since wider regions require a higher number of seeds, while noises (i.e., small components) should not contain any, the number of internal seeds k i is proportional to the size of its respective component C i .…”

Section: A Seed Samplingmentioning

confidence: 99%

“…The user can establish the degree of such penalization through an importance factor γ. Since the previous work deals with binary images and saliency maps might not be such, we adapted the connectivity function [8] in order to penalize the saliency transitions along the path. Thus, higher importance values impose a harsher penalization, causing an exponential growth of the path-costs and, consequently, reducing the chances of superpixel leaking.…”

Section: B Superpixel Generationmentioning

confidence: 99%

“…In this dissertation, we propose a generalization of the Iterative Spanning Forest (ISF) [7] framework, which allows the inclusion of prior object information, represented as object saliency maps -i.e., monochromatic images whose pixel value indicates the proportional likelihood of belonging to an object of interest. Named Object-based ISF (OISF) [8], our framework maintains the properties of ISF, and straightforwardly incorporates the saliency information in each step of the pipeline. OISF-based methods can alter the superpixel resolution within the objects of interest, whilst permitting a better delineation of the object (Figure 1); being a suitable tool for distinct applications, such as feature extraction and semantic segmentation.…”

Section: Introductionmentioning

confidence: 99%

“…Note that such improvement can be achieved without the need to increase the number of superpixels. Moreover, we also propose two object-based seed sampling strategies: the Object Geodesic Grid Sampling (OGRID) [8], and the Object Saliency Map by Ordered eXtraction (OS-MOX) [9]. As stated in [9], the simple consideration of one of both leads to a significant improvement for object delineation.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Superpixel Generation by the Iterative Spanning Forest Using Object Information

Belém

Falcão

Guimarães

2020

Anais Estendidos Da Conference on Graphics, Patterns and Images (SIBRAPI Estendido 2020)

View full text Add to dashboard Cite

Superpixel segmentation methods aim to partition the image into homogeneous connected regions of pixels (i.e., superpixels) such that the union of its comprising superpixels precisely defines the objects of interest. However, the homogeneity criterion is often based solely on color, which, in certain conditions, might be insufficient for inferring the extension of the objects (e.g., low gradient regions). In this dissertation, we address such issue by incorporating prior object information — represented as monochromatic object saliency maps — into a state-of-the-art method, the Iterative Spanning Forest (ISF) framework, resulting in a novel framework named Object-based ISF (OISF). For a given saliency map, OISF-based methods are capable of increasing the superpixel resolution within the objects of interest, whilst permitting a higher adherence to the map’s borders, when color is insufficient for delineation. We compared our work with state-of-the-art methods, considering two classic superpixel segmentation metrics, in three datasets. Experimental results show that our approach presents effective object delineation with a significantly lower number of superpixels than the baselines, especially in terms of preventing superpixel leaking.

show abstract

Section: Object-based Superpixel Segmentationmentioning

confidence: 99%

Section: A Seed Samplingmentioning

confidence: 99%

Section: B Superpixel Generationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%