{"title":"基于纹理的SLIC超像素高空间分辨率遥感图像分割方法","authors":"Lizhen Lu, Chuan Wang, Xiao Yin","doi":"10.1109/Agro-Geoinformatics.2019.8820692","DOIUrl":null,"url":null,"abstract":"Super-pixel methods cluster spatially connected similar pixels into perceptually meaningful regions, which are generally used as basic units instead of the original pixels in pre-processing and segmentation of high spatial resolution images for the object-oriented image classification. Among a number of super-pixel methods, the simple linear iterative clustering (SLIC) has been widely applied due to its simplicity, efficiency, and ability to adhere to image boundaries. SLIC itself, however, was originally designed to group black-white or three-color common images rather than multi-spectral/ hyperspectral remote sensing ones into super-pixels. In order to better apply SLIC to segmenting remote sensing images at high spatial resolution, the SLIC algorithm was modified by incorporating grey-level co-occurrence matrix texture with color features and expanding measure approach for weighted distance of texture and color similarity and spatial proximity between super-pixel center and neighboring pixels. Gaofen-2 panchromatic, multispectral and fused images were used to valid the modified SLIC (MSLIC) algorithm. Both completeness (CPS) and correctness (CRS) were used to quantitatively evaluate both MSLIC and SLIC algorithms. Visually interpreting approach was also applied to compare the segmentation and classification maps from the two algorithms. The experimental results indicate MSLIC achieves higher CPS and CRS than SLIC.","PeriodicalId":143731,"journal":{"name":"2019 8th International Conference on Agro-Geoinformatics (Agro-Geoinformatics)","volume":null,"pages":null},"PeriodicalIF":0.0000,"publicationDate":"2019-07-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"5","resultStr":"{\"title\":\"Incorporating Texture into SLIC Super-pixels Method for High Spatial Resolution Remote Sensing Image Segmentation\",\"authors\":\"Lizhen Lu, Chuan Wang, Xiao Yin\",\"doi\":\"10.1109/Agro-Geoinformatics.2019.8820692\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Super-pixel methods cluster spatially connected similar pixels into perceptually meaningful regions, which are generally used as basic units instead of the original pixels in pre-processing and segmentation of high spatial resolution images for the object-oriented image classification. Among a number of super-pixel methods, the simple linear iterative clustering (SLIC) has been widely applied due to its simplicity, efficiency, and ability to adhere to image boundaries. SLIC itself, however, was originally designed to group black-white or three-color common images rather than multi-spectral/ hyperspectral remote sensing ones into super-pixels. In order to better apply SLIC to segmenting remote sensing images at high spatial resolution, the SLIC algorithm was modified by incorporating grey-level co-occurrence matrix texture with color features and expanding measure approach for weighted distance of texture and color similarity and spatial proximity between super-pixel center and neighboring pixels. Gaofen-2 panchromatic, multispectral and fused images were used to valid the modified SLIC (MSLIC) algorithm. Both completeness (CPS) and correctness (CRS) were used to quantitatively evaluate both MSLIC and SLIC algorithms. Visually interpreting approach was also applied to compare the segmentation and classification maps from the two algorithms. The experimental results indicate MSLIC achieves higher CPS and CRS than SLIC.\",\"PeriodicalId\":143731,\"journal\":{\"name\":\"2019 8th International Conference on Agro-Geoinformatics (Agro-Geoinformatics)\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2019-07-16\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"5\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2019 8th International Conference on Agro-Geoinformatics (Agro-Geoinformatics)\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/Agro-Geoinformatics.2019.8820692\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2019 8th International Conference on Agro-Geoinformatics (Agro-Geoinformatics)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/Agro-Geoinformatics.2019.8820692","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Incorporating Texture into SLIC Super-pixels Method for High Spatial Resolution Remote Sensing Image Segmentation
Super-pixel methods cluster spatially connected similar pixels into perceptually meaningful regions, which are generally used as basic units instead of the original pixels in pre-processing and segmentation of high spatial resolution images for the object-oriented image classification. Among a number of super-pixel methods, the simple linear iterative clustering (SLIC) has been widely applied due to its simplicity, efficiency, and ability to adhere to image boundaries. SLIC itself, however, was originally designed to group black-white or three-color common images rather than multi-spectral/ hyperspectral remote sensing ones into super-pixels. In order to better apply SLIC to segmenting remote sensing images at high spatial resolution, the SLIC algorithm was modified by incorporating grey-level co-occurrence matrix texture with color features and expanding measure approach for weighted distance of texture and color similarity and spatial proximity between super-pixel center and neighboring pixels. Gaofen-2 panchromatic, multispectral and fused images were used to valid the modified SLIC (MSLIC) algorithm. Both completeness (CPS) and correctness (CRS) were used to quantitatively evaluate both MSLIC and SLIC algorithms. Visually interpreting approach was also applied to compare the segmentation and classification maps from the two algorithms. The experimental results indicate MSLIC achieves higher CPS and CRS than SLIC.