Bayesian Adaptive Superpixel Segmentation

Uziel, Roy; Ronen, Meitar; Freifeld, Oren

doi:10.1109/iccv.2019.00856

Cited by 37 publications

(33 citation statements)

References 43 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The parameters of this unknown distribution are obtained by maximum likelihood estimation, which are then used to estimate the superpixel labels based on posterior probability. Another similar method coined as Bayesian Adaptive Superpixel Segmentation (BASS) [13], based on the Dirichlet-Process Gaussian Mixture Model (DPGMM), aims to discover the latent distribution of superpixels, adapting from an initial K, as well as the distribution z i that represents the measurement-to-cluster assignment, i.e. the label of pixel x i .…”

Section: Related Workmentioning

confidence: 99%

“…The quality of the segmentation of IBIS is evaluated and compared to 7 state of the art methods SLIC [7], GMMSP [12], SEEDS [6], LSC [22], ERS [10], USEQ [5] and BASS [13]. SLIC serves as a reference for us since IBIS inherits the pixel aggregation, chromatic and spatial distance calculation from it.…”

Section: A Evaluation Of Ibis Segmentation Qualitymentioning

confidence: 99%

“…FIGURE13: Evolution of calculation times for IBIScuda (on the left) and gSLICr (on the right), on the Jetson TX1 GPU platform, depending on the pixel resolution of the image for superpixel resolutions 50, 100, 250, 500, 1000 and 2500.…”

mentioning

confidence: 99%

See 2 more Smart Citations

Iterative Boundaries Implicit Identification for Superpixels Segmentation: A Real-Time Approach

Bobbia¹,

Macwan²,

Benezeth³

et al. 2021

IEEE Access

View full text Add to dashboard Cite

Superpixel algorithms group visually coherent pixels and form an alternative representation of the regular structure of the pixel grid. This fundamental low-level computer vision preprocessing step greatly reduces the complexity of subsequent image processing tasks. However, most of the existing methods suffer from very high calculation costs which makes them quite unsuitable for time-sensitive applications. In this paper, we propose a new superpixel segmentation method, named IBIS for Iterative Boundaries implicit Identification for superpixels segmentation, that implicitly identifies the boundaries between superpixels and performs the segmentation using only a fraction of the pixels of the input image, thereby reducing the complexity and computation time. The results obtained during the experiments show that the segmentation quality of IBIS is comparable to that of state of the art methods with a computation time divided by a factor of 8 without parallelization of the processing for low resolution images (e.g., 320x240 pixels) as usually provided in public data sets. We also present and comprehensively evaluate the GPU variant of IBIS named IBIScuda that allows an optimal exploitation of the available resources considering the limited bandwidth between CPU and GPU memories.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: A Evaluation Of Ibis Segmentation Qualitymentioning

confidence: 99%

See 1 more Smart Citation

Iterative Boundaries Implicit Identification for Superpixels Segmentation: A Real-Time Approach

Bobbia¹,

Macwan²,

Benezeth³

et al. 2021

IEEE Access

View full text Add to dashboard Cite

show abstract

“…The adaptive algorithm can use the feedback values from previous iterations to dynamically adjust the parameters for better performance. This type of solutions has been proposed for adapting the used weights for segmentation or graph optimization in 2D superpixel segmentation algorithms with promising results, e.g., [22], [23]. Given the similarities between 2D and 4D LF image segmentation, a similar approach can be followed for LF images.…”

Section: Introductionmentioning

confidence: 99%

ALFO: Adaptive Light Field Over-Segmentation

Hamad¹,

Conti

Nunes

et al. 2021

IEEE Access

View full text Add to dashboard Cite

Automatic image over-segmentation into superpixels has attracted increasing attention from researchers to apply it as a pre-processing step for several computer vision applications. In 4D Light Field (LF) imaging, image over-segmentation aims at achieving not only superpixel compactness and accuracy but also cross-view consistency. Due to the high dimensionality of 4D LF images, depth information can be estimated and exploited during the over-segmentation along with spatial and visual appearance features. However, balancing between several hybrid features to generate robust superpixels for different 4D LF images is challenging and not adequately solved in existing solutions. In this paper, an automatic, adaptive, and view-consistent LF over-segmentation method based on normalized LF cues and K-means clustering is proposed. Initially, disparity maps for all LF views are estimated entirely to improve superpixel accuracy and consistency. Afterwards, by using K-means clustering, a 4D LF image is iteratively divided into regular superpixels that adhere to object boundaries and ensure cross-view consistency. Our proposed method can automatically adjust the clustering weights of the various features that characterize each superpixel based on the image content. Quantitative and qualitative results on several 4D LF datasets demonstrate outperforming performance of the proposed method in terms of superpixel accuracy, shape regularity and view consistency when using adaptive clustering weights, compared to the state-of-the-art 4D LF over-segmentation methods.INDEX TERMS Automatic segmentation, adaptive light field over-segmentation, superpixels.

show abstract

“…These three advantages of superpixel motivates our interest in superpixel proposal generation. In the realm of precisely capturing the irregular object boundary, we can at least compare with two other popular techniques such as adaptive segmentation [89,90], and attention. In more general terms, there exist model-free methods and model-based (or learnable) methods.…”

Section: Deep Superpixel Neural Network (Dspnn)mentioning

confidence: 99%

Deep heterogeneous superpixel neural networks for image analysis and feature extraction

Yang¹

View full text Add to dashboard Cite

Lately, deep convolutional neural networks are rapidly transforming and enhancing computer vision accuracy and performance, and pursuing higher-level and interpretable object recognition. Superpixel-based methodologies have been used in conventional computer vision research where their efficient representation has superior effects. In contemporary computer vision research driven by deep neural networks, superpixel-based approaches mainly rely on oversegmentation to provide a more efficient representation of the imagery data, especially when the computation is too expensive in time or memory to perform in pairwise similarity regularization or complex graphical probabilistic inference. In this dissertation, we proposed a novel superpixel-enabled deep neural network paradigm by relaxing some of the prior assumptions in the conventional superpixel-based methodologies and exploring its capabilities in the context of advanced deep convolutional neural networks. This produces novel neural network architectures that can achieve higher-level object relation modeling, weakly supervised segmentation, high explainability, and facilitate insightful visualizations. This approach has the advantage of being an efficient representation of the visual signal and has the capability to dissect out relevant object components from other background noise by spatially re-organizing visual features. Specifically, we have created superpixel models that join graphical neural network techniques and multiple-instance learning to achieve weakly supervised object detection and generate precise object bounding without pixel-level training labels. This dissection and the subsequent learning by the architecture promotes explainable models, whereby the human users of the models can see the parts of the objects that have led to recognition. Most importantly, this neural design's natural result goes beyond abstract rectangular bounds of an object occurrence (e.g., bounding box or image chip), but instead approaches efficient parts-based segmented recognition. It has been tested on commercial remote sensing satellite imagery and achieved success. Additionally, We have developed highly efficient monocular indoor depth estimation based on superpixel feature extraction. Furthermore, we have demonstrated state-of-theart weakly supervised object detection performance on two contemporary benchmark data sets, MS-COCO and VOC 2012. In the future, deep learning techniques based on superpixel-enabled image analysis can be further optimized in accuracy and computational performance; and it will also be interesting to evaluate in other research domains, such as those involving medical imagery, infrared imagery, or hyperspectral imagery.

show abstract

Bayesian Adaptive Superpixel Segmentation

Cited by 37 publications

References 43 publications

Iterative Boundaries Implicit Identification for Superpixels Segmentation: A Real-Time Approach

Iterative Boundaries Implicit Identification for Superpixels Segmentation: A Real-Time Approach

ALFO: Adaptive Light Field Over-Segmentation

Deep heterogeneous superpixel neural networks for image analysis and feature extraction

Contact Info

Product

Resources

About