Improvement of Image Segmentation Accuracy Based on Multiscale Optimization Procedure

Abstract

This letter proposes an optimization approach that enhances the quality of image segmentation using the software Definiens Developer. The procedure aims at the minimization of over-and undersegmentations in order to attain more accurate segmentation results. The optimization iteratively combines a sequence of multiscale segmentation, feature-based classification, and classification-based object refinement. The developed method has been applied to various remotely sensed data and is compared to the results achieved with the established segmentation procedures provided by the Definiens Developer software. The quantitative assessment of segmentation accuracy based on reference objects is derived from an aerial image, and a high-resolution synthetic aperture radar scene shows an improvement of 20%-40% in object accuracy by applying the proposed procedure.Index Terms-High-resolution data, image segmentation, level fusion, object refinement, object-oriented image analysis.

Object-based integrated landscape change analysis: synergy of multi-temporal LiDAR and very high resolution orthophotos

“…The parameter settings used for scale parameter are 10, 20, 30, 40 and 50, for the shape parameter 0.1, 0.3, 0.5, 0.7, 0.9 and for compactness 0.1, 0.3, 0.5, 0.7 and 0.9. The accuracy of the segmentation results was determined by comparing the generated objects with the orthophotos based on over-and under segmentation (Esch et al, 2008). Based on the accuracy assessment a MRS using a scale parameter of 50, a shape parameter of 0.3 and compactness of 0.5, and the spectral information from the orthophotos in combination with the LiDAR-derived canopy height model (CHM) was used.…”

Vision Goes Symbolic Without Loss of Information Within the Preattentive Vision Phase: The Need to Shift the Learning Paradigm from Machine-Learning (from Examples) to Machine-Teaching (by Rules) at the First Stage of a Two-Stage Hybrid Remote... Part II

“…Due to the availability of a commercial GEOBIA software developed by a German company (Definiens Imaging GmbH, 2004;Esch et al, 2008), two-stage segment-based RS-IUSs have recently gained widespread popularity and are currently considered the state-of-the-art in both scientific and commercial RS image mapping application domains (Mather, 1994;Pekkarinen, Reithmaier & Strobl, 2009). In practice, under the guise of 'flexibility' current commercial 2-D object-based software provides overly complicated options to choose from (Hay & Castilla, 2006).…”

“…However, it differs from the CV system proposed by Marr at the level of primal sketch implementation (see Part I Section 2.6) consisting of a sub-symbolic zero-crossing algorithm (Marr, 1982). In addition, the novel RS-IUS sketched above differs at the level of both computational theory and algorithm design and implementation from existing CV systems such as GEOBIA systems (Definiens Imaging GmbH, 2004;Esch et al, 2008), including Diamant's (Diamant, 2005;Diamant, 2008;Diamant, 2010a;Diamant, 2010b), where an unlabeled data learning (driven-without-knowledge) algorithm is adopted at the first stage.…”

Remote Sensing-Based Characterization of Settlement Structures for Assessing Local Potential of District Heat

“…This building mask is integrated into the second analysis step where the optical data are first segmented into image objects by means of an image segmentation optimization workflow developed by [52] and then classified based on a method presented by [53]. The result of the combined analysis of spectral information and height information is a detailed LU/LC classification of the covered area of Oberhaching.…”