Color Texture Segmentation Based on the Modal Energy of Deformable Surfaces

Abstract

This paper presents a new approach for the segmentation of color textured images, which is based on a novel energy function. The proposed energy function, which expresses the local smoothness of an image area, is derived by exploiting an intermediate step of modal analysis that is utilized in order to describe and analyze the deformations of a 3-D deformable surface model. The external forces that attract the 3-D deformable surface model combine the intensity of the image pixels with the spatial information of local image regions. The proposed image segmentation algorithm has two steps. First, a color quantization scheme, which is based on the node displacements of the deformable surface model, is utilized in order to decrease the number of colors in the image. Then, the proposed energy function is used as a criterion for a region growing algorithm. The final segmentation of the image is derived by a region merge approach. The proposed method was applied to the Berkeley segmentation database. The obtained results show good segmentation robustness, when compared to other state of the art image segmentation algorithms.

A Precision-Recall Criterion Based Consensus Model For Fusing Multiple Segmentations

“…We have also tested the performance of our fusion method as segmentation method, [12] 0.81 -2011-gPb-owt-ucm [64] 0.81 -2012-AMUS [52] 0.80 -2010-PRIF [53] 0.80 -2008-CTex [5] 0.80 -2009-MIS [36] 0.80 -2011-SCKM [7] 0.80 -2008-FCR [4] 0.79 -2012-SFSBM [62] 0.79 -2004-FH [30] (in [6]) 0.78 -2011-MD2S [11] 0.78 -2009-HMC [25] 0.78 -2009-Consensus [50] 0.78 -2009-Total Var. [33] 0.78 -2009-A-IFS HRI [10] 0.77 -2001-JSEG [17] (in [5]) 0.77 -2011-KM [34] 0.76 -2007-CTM [6,18] 0 in term of F measure (see Table 3).…”

“…Years of research in segmentation have thus focused on finding more sophisticated image features and/or more elaborate clustering techniques and significant improvements in the final segmentation results have been achieved, generally at the cost of an increase in model complexity and/or in computational complexity. These methods include segmentation models exploiting directly clustering schemes [4,5,6,7] using Gaussian mixture modeling, fuzzy clustering approach [8,9] or fuzzy sets [10] or after a possibly de-texturing approach [7,11,12]), mean-shift or more generally mode seeking based procedures [13,14,15], watershed or [16] region growing strategies [17], lossy coding and compression models [18,16], wavelet transform [19], MRF [20,21,22,23,24,25,26], Bayesian [27] texton-based approach [28] or graph-based models [29,30,31], variational or level set methods [32,33,34,28,35], deformable surfaces [36], active contour model [37] (with graph partitioning based approach [38]) or curve-based techniques, iterative unsupervised thresholding technique [39,40], genetic algorithm …”

MDS-Based Multiresolution Nonlinear Dimensionality Reduction Model for Color Image Segmentation

“…-2010-PRIF [29] 0.801 -2008-CTex [18] 0.800 -2009-MIS [13] 0.798 -2008-FCR [17] 0.788 -2004-FH [8] (in [19]) 0.784 -2009-HMC [3] 0.783 -2009-Consensus [30] 0.781 -2009-A-IFS HRI [21] 0.771 -2001-JSEG [26] (in [18]) 0.770 -2007-CTM [19] 0.762 -2008-St-SVGMM [22] 0.759 -2003-Mean-Shift [4] (in [19]) 0.755 -2008-NTP [9] 0.752 -2010-iHMRF [4] 0.752 -2006-GBMS [16] (in [6]) 0.734 -2000-NCuts [7] (in [19]) 0.722 -2010-JND [24] 0.719 -2010-DCM [6] 0.708 -2009- [25] 0.703…”

“…Digital Object Identifier 10.1109/TNN.2010.2101614 [10], [11], variational methods [12], deformable surfaces [13], mean-shift-based techniques [14]- [16], clustering schemes [17]- [20] (with fuzzy sets [21] or Gaussian mixture models [22], [23]), color histograms [24], watershed techniques [25], region growing strategies [26], region-based split and merge procedures (sometimes directly expressed by a global energy function to be optimized [27]), and finally some fusion methods [17], [28]- [30]. Most of these above-mentioned methods preliminarily use (sometimes conjointly to the segmentation step) a texture feature extraction step whose goal is to characterize each meaningful textured region (to be segmented) with statistical (or geometrical, morphological, fractal, etc.,) image features which are then either characterized by their distribution or simply gathered in a D-dimensional (feature) vector.…”

Integration of color and texture cues in a rough set–based segmentation method

“…Examples of these methods include the split-and-merge, 4,5 region growing, 6,7 watershed, 8,9 and energy minimization. [10][11][12] The main drawback of such approaches is that, even though the resulting regions are spatially well connected, there is no guarantee that the segments are homogeneous for a specific feature space. Moreover, the pixelby-pixel agglomeration strategy of these procedures often results in intensive computational schemes.…”