Quasi-flat Zones Research Articles

Hierarchical multi-label propagation using speaking face graphs for multimodal person discovery

TV archives are growing in size so fast that manually indexing becomes unfeasible. Automatic indexing techniques can be applied to overcome this issue, and this work proposes an unsupervised technique for multimodal person discovery. To achieve this goal, we propose a hierarchical label propagation technique based on quasi-flat zones theory, that learns from labeled and unlabeled data and propagates names through a multimodal graph representation. In this representation, we combine audio, video, and text processing techniques to model the data as a graph of speaking faces. In the proposed mod-eling, we extract names via optical character recognition and propagate them through the graph using audiovisual relationships between speaking faces. We also use a random walk label propagation and two graph clustering strategies to serve as baselines. The proposed label propagation techniques always outper-form the clustering baselines on the quantitative assessments. Our approach also outperforms all literature methods tested on the same dataset except for one, which uses a different preprocessing step. The proposed hierarchical label propagation and the random walk baseline produce highly equivalent results according to the Kappa coefficient, but the hierarchical propagation is parameter-free and over 9 times faster than the random walk under the same configurations.

Color Image Complexity versus Over-Segmentation: A Preliminary Study on the Correlation between Complexity Measures and Number of Segments.

It is said that image segmentation is a very difficult or complex task. First of all, we emphasize the subtle difference between the notions of difficulty and complexity. Then, in this article, we focus on the question of how two widely used color image complexity measures correlate with the number of segments resulting in over-segmentation. We study the evolution of both the image complexity measures and number of segments as the image complexity is gradually decreased by means of low-pass filtering. In this way, we tackle the possibility of predicting the difficulty of color image segmentation based on image complexity measures. We analyze the complexity of images from the point of view of color entropy and color fractal dimension and for color fractal images and the Berkeley data set we correlate these two metrics with the segmentation results, more specifically the number of quasi-flat zones and the number of JSEG regions in the resulting segmentation map. We report on our experimental results and draw conclusions.

Levellings based on spatially adaptive scale spaces using local image features

The authors propose here to overcome lacks of robustness against noise and adaptability to image features for which classical morphological operators suffer from. For doing this, they propose to deal with partial differential equations (PDEs) for generalised Cauchy problems, and they show that the proposed PDEs are equivalent to impose both robustness and adaptability to structuring functions of the corresponding sup-inf operators. This allows them to introduce spatially adaptability in levellings, and it turns out that the proposed approach constitutes a PDE formulation and a generalisation of a larger class of levellings, the so-called extended levellings, for which one of them are characterised by quasi-flat zones. They show the efficiency of the proposed approach on synthetic, grey, and colour images with different types of noises.

Reducing the oversegmentation induced by quasi-flat zones for multivariate images

Quasi-flat zones are morphological operators which partition the image into homogeneous regions with respect to certain criteria. They are used in grayscale and multivariate image simplification and segmentation. However, they often induce an oversegmentation of the image, taking the shape of narrow transition regions between objects and small regions which are a few pixels wide. Various methods have been devised in order to reduce this oversegmentation, which remove the unwanted zones according to some criteria and then grow the remaining regions. In this paper we propose improvements in transition region and area threshold filtering. We also combine the two filtering methods for further-improved results. We apply the proposed approaches in color image segmentation and hyperspectral pixel classification.

Evaluation of Hierarchical Watersheds.

This paper aims to understand the practical features of hierarchies of morphological segmentations, namely the quasi-flat zones hierarchy and watershed hierarchies, and to evaluate their potential in the context of natural image analysis. We propose a novel evaluation framework for the hierarchies of partitions designed to capture various aspects of those representations: precision of their regions and contours, possibility to extract high quality horizontal cuts and optimal non-horizontal cuts for image segmentation, and the ease of finding a set of regions representing a semantic object. This framework is used to assess and to optimize hierarchies with respect to the possible pre- and post-processing steps. We show that, used in conjunction with a state-of-the-art contour detector, watershed hierarchies are competitive with the complex state-of-the-art methods for hierarchy construction. In particular, the proposed framework allows us to identify a watershed hierarchy based on a novel extinction value, the number of parent nodes that outperforms the other hierarchies of morphological segmentations. This coupled with the fact that watershed hierarchies satisfy clear global optimality properties and can be efficiently computed on large data, make them valuable candidates for various computer vision tasks.

Hierarchical Segmentations with Graphs: Quasi-flat Zones, Minimum Spanning Trees, and Saliency Maps

Hierarchies of partitions are generally represented by dendrograms (direct representation). They can also be represented by saliency maps or minimum spanning trees. In this article, we precisely study the links between these three representations. In particular, we provide a new bijection between saliency maps and hierarchies based on quasi-flat zones as often used in image processing and we characterize saliency maps and minimum spanning trees as solutions to constrained minimization problems where the constraint is quasi-flat zones preservation. In practice, these results make up a toolkit for designing new hierarchical methods where one can choose the most convenient representation. They also invite us to process non-image data with morphological hierarchies. More precisely, we show the practical interest of the proposed framework for: i) hierarchical watershed image segmentations, ii) combinations of dierent hierarchical segmentations, iii) hierarchicalizations of some non-hierarchical image segmentation methods based on regional dissimilarities, and iv) hierarchical analysis of geographical data.

On Attribute Thresholding and Data Mapping Functions in a Supervised Connected Component Segmentation Framework

Search-centric, sample supervised image segmentation has been demonstrated as a viable general approach applicable within the context of remote sensing image analysis. Such an approach casts the controlling parameters of image processing—generating segments—as a multidimensional search problem resolvable via efficient search methods. In this work, this general approach is analyzed in the context of connected component segmentation. A specific formulation of connected component labeling, based on quasi-flat zones, allows for the addition of arbitrary segment attributes to contribute to the nature of the output. This is in addition to core tunable parameters controlling the basic nature of connected components. Additional tunable constituents may also be introduced into such a framework, allowing flexibility in the definition of connected component connectivity, either directly via defining connectivity differently or via additional processes such as data mapping functions. The relative merits of these two additional constituents, namely the addition of tunable attributes and data mapping functions, are contrasted in a general remote sensing image analysis setting. Interestingly, tunable attributes in such a context, conjectured to be safely useful in general settings, were found detrimental under cross-validated conditions. This is in addition to this constituent’s requiring substantially greater computing time. Casting connectivity definitions as a searchable component, here via the utilization of data mapping functions, proved more beneficial and robust in this context. The results suggest that further investigations into such a general framework could benefit more from focusing on the aspects of data mapping and modifiable connectivity as opposed to the utility of thresholding various geometric and spectral attributes.

The impact of multivariate quasi-flat zones on the morphological description of hyperspectral images

Quasi-flat zones are powerful morphological image simplification tools capable of producing unique image partitions at multiple scales. In this paper we investigate their impact on the morphological description of hyperspectral images and, specifically, whether they can improve classification performance when used for preprocessing. In particular, we employ them in order to group spatially adjacent pixels according to spectral criteria, prior to the computation of extended morphological profiles. Moreover, we explore both marginal quasi-flat zones and recent multivariate extensions using well-known vector orderings. The method studied is tested with multiple hyperspectral images, where it leads consistently to improvements in classification performance.

Fast quasi-flat zones filtering using area threshold and region merging

Quasi-flat zones are morphological operators which segment the image into homogeneous regions according to certain criteria. They are used as an image simplification tool or an image segmentation pre-processing, but they induced a very important oversegmentation. Several filtering methods have been proposed to deal with this issue but they suffer from different drawbacks, e.g., loss of quality or edge deformation. In this article, we propose a new method based on existing approaches which achieves better or similar results than existing approaches, does not suffer from their drawbacks and requires less computation time. It consists of two successive steps. First, small quasi-flat zones are removed according to a minimal area threshold. They are then filled through the growth of remaining zones.

Morphological hierarchical segmentation and color spaces

AbstractIn this article, the color space influence on different methods of morphological hierarchical segmentation is studied. For this purpose, several experimental studies are carried out on hundred images of the “Berkeley Segmentation Dataset and Benchmark” (Berkeley). On one hand, four usual and representative color spaces (RGB, YCbCr, L*a*b*, IHSL) are considered. On the other hand, three different morphological hierarchical segmentation schemes are considered. These later are the quasi flat zones, the nonparametric hierarchical watershed (waterfall algorithm) and the nested homogeneous connections. As one could expect it, the different morphological hierarchical segmentation schemes do not have the same behavior in different color spaces. We conclude by proposing a ranking of color spaces according to the considered morphological hierarchical segmentation schemes. © 2010 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 20, 167–178, 2010; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ima.20233

Iterative area filtering of multichannel images

Owing to the absence of total ordering between vectors of more than one dimension, most morphological transformations are not directly applicable to multichannel images. In this paper, we show that area based operators can be extended to the processing of multichannel images by using the quasi-flat zones and seeded region growing paradigms. The method is iterative in the sense that it starts with the smallest non-unitary quasi-flat zones until an automatically derived contrast threshold value is reached. Pseudo-code is given for all algorithms. We illustrate their usefulness for the simplification of complex natural images such as those occurring in satellite images of the earth.