Fast and Effective Superpixel Segmentation Using Accurate Saliency Estimation

Belém, Felipe; Borlido, Isabela; João, Leonardo; Cousty, Jean; Guimarães, Silvio Jamil Ferzoli; Falcão, Alexandre X.

doi:10.1007/978-3-031-19897-7_21

Cited by 7 publications

(6 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…• novel arc-cost functions which incorporate prior object information and produce compact and irregular superpixels robust to saliency errors; • novel seed relevance criterion for superpixel spreading and three novel seed relevance saliency-based penalization for improving delineation and assuring overall compactness; • thorough optimization of SICLE and an extensive performance evaluation in five datasets, considering well-known metrics and seven stateof-the-art methods as baselines, resulting in two SICLE variants for effective delineation and compact superpixel generation; • debate over multiobject handling in SICLE. This article extends and completes a previous work published at a conference [21]. Differently from it, the saliency importance in our arc-cost functions is not a binary parameter (thus, may be optimized), and the seed penalization is no longer associated with such importance, being independently defined.…”

Section: Introductionsupporting

confidence: 60%

“…The speed performance of the fastest methods in ECSSD is presented in Table 2. It is important to state that the performance of SICLE-IRREG differs from the results shown in [21]. We argue that the speedup achieved from exploiting DIFTs is limited when a significantly high quantity of seeds is removed between iterations, which is a direct result when Ω = 2.…”

Section: Object-based Analysismentioning

confidence: 79%

“…In this work, our goal is to provide SICLE variants for generating effective delineation with irregular and compact superpixels (hereafter named SICLE-IRREG and SICLE-COMP, respectively). Moreover, differently from [20,21], these novel versions of SICLE exploit differential computations of the IFT (DIFTs) [45] for improving speed. For comparison, we selected the following state-of-theart algorithms as baselines based on their reported performance: (i) ERS [31] 1 ; (ii) SLIC [7] 2 ; (iii) SH [9] 3 ; (iv) IBIS [8] 4 ; (v) LSC [24] 5 ; (vi) GMM [28] 6 ; (vii) OISF [17,18] 7 ; and (viii) DAL-HERS [36] 8 .…”

Section: Methodsmentioning

confidence: 99%

“…Differently from it, the saliency importance in our arc-cost functions is not a binary parameter (thus, may be optimized), and the seed penalization is no longer associated with such importance, being independently defined. The remaining contributions are original and do not appear in [21].…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Novel Arc-Cost Functions and Seed Relevance Estimations for Compact and Accurate Superpixels

et al. 2023

Self Cite

View full text Add to dashboard Cite

On the verge of superpixel methods exploiting saliency information, the Superpixels through Iterative CLEarcutting (SICLE) framework has reported fast and accurate superpixel delineation. It is composed of three steps: (i) seed oversampling; (ii) superpixel generation; and (iii) seed removal. It starts from (i) and applies several iterations of (ii) and (iii) until reaching the desired superpixel quantity. In this work, we improve SICLE such that it can now generate compact superpixels with accurate delineation. We exploit differential computation and propose several novel functions for steps (ii) and (iii) for proper saliency incorporation, compact superpixel generation, and improvement in speed and delineation. Results show that, with our proposals, SICLE achieves state-of-the-art performance in delineation and speed whenever saliency is absent with on-par compacity. When an accurate saliency map is provided, its performance improves significantly and requires only two iterations for segmentation.

show abstract

Section: Introductionsupporting

confidence: 60%

Section: Object-based Analysismentioning

confidence: 79%

Section: Methodsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Novel Arc-Cost Functions and Seed Relevance Estimations for Compact and Accurate Superpixels

et al. 2023

Self Cite

View full text Add to dashboard Cite

show abstract

“…It also has an international conference paper [40], and a journal under review in ACM Surveys. Another conference paper [41] was published in parallel but is not part of the dissertation.…”

Section: Discussionmentioning

confidence: 99%

A survey on the state-of-the-art superpixel segmentation

Barcelos,

Falcão,

Guimarães

2023

Anais Estendidos Da XXXVI Conference on Graphics, Patterns and Images (SIBRAPI Estendido 2023)

View full text Add to dashboard Cite

In contrast to the rapid progress of superpixel segmentation, their methods are often compared only with classical approaches. Also, the previous superpixel methods categorizations are insufficient to cover the recent literature. In addition, although the inner color similarity usually underlies superpixel methods, both color homogeneity measures have important drawbacks. In this work, we fill these gaps by providing a new taxonomy for superpixel segmentation, a new color homogeneity measure, and an extensive comparison among 20 superpixel methods. Experiments show that the proposed measure, named Similarity between Image and Reconstruction from Superpixels (SIRS), is more robust to slight color variations than Explained Variation. Using SIRS and the commonly used superpixel metrics, we evaluated 20 superpixel segmentation methods and provided insights into the different approaches based on the clustering categories in our taxonomy.

show abstract