Kwis
ÓPTICA PURA Y APLICADA. www.sedoptica.es
Color Space Analysis for Color Image Segmentation of Natural Scenes by K-means Based Clustering
Jia Song(1,∗) , Eva M. Valero(1) , Juan L. Nieves(1)
1. Optics Department, Faculty of Science, University of Granada, Spain
(∗)
Corresponding author Email: songjia815@gmail.com
Recibido / Received: dd/mm/yyyy Aceptado / Accepted: dd/mm/yyyy
ABSTRACT: In this paper, we propose to segment color images of natural scenes by pixel clustering in different color spaces in order to compare the performances of the color spaces. Eight color spaces are evaluated, including perceptual and non-perceptual color spaces. The clustering methods applied in this work are k-means and its extensions. Segmentation results are evaluated by calculating the degree of matching with the benchmark created by manual labeling, and visual evaluation is also applied as a reference. A series of experiments were designed and conducted. Results show that the performances of the color spaces are dependent on the scenes, and it is possible to choose a proper color space manually for a specific image scene segmentation task according to the data dependency patterns analyzed from the experiment results. Key words: color image segmentation, natural scenes, color spaces, k-means clustering
References
[1] Busin, L., Vandenbroucke, N., and Macaire, L., "Color spaces and image segmentation", Advances in Imaging and Electron Physics 151, 65–168 (2008). [2] Cheng, H., Jiang, X., Sun, Y., and Wang, J., "Color image segmentation: advances and prospects", Pattern Recognition 34(12), 2259–2281 (2001). [3] Busin, L., Shi, J., Vandenbroucke, N., and Macaire, L., "Color space selection for color image segmentation by spectral clustering", IEEE International Conference on Signal and Image Processing Applications 22(8), 262–267 (2009). [4] Fernando E. Correa-Tome, Raul E. Sanchez-Yanez and Victor Ayala-Ramirez, "Comparison of perceptual color spaces for natural image
Color Space Analysis for Color Image Segmentation of Natural Scenes by K-means Based Clustering
Jia Song(1,∗) , Eva M. Valero(1) , Juan L. Nieves(1)
1. Optics Department, Faculty of Science, University of Granada, Spain
(∗)
Corresponding author Email: songjia815@gmail.com
Recibido / Received: dd/mm/yyyy Aceptado / Accepted: dd/mm/yyyy
ABSTRACT: In this paper, we propose to segment color images of natural scenes by pixel clustering in different color spaces in order to compare the performances of the color spaces. Eight color spaces are evaluated, including perceptual and non-perceptual color spaces. The clustering methods applied in this work are k-means and its extensions. Segmentation results are evaluated by calculating the degree of matching with the benchmark created by manual labeling, and visual evaluation is also applied as a reference. A series of experiments were designed and conducted. Results show that the performances of the color spaces are dependent on the scenes, and it is possible to choose a proper color space manually for a specific image scene segmentation task according to the data dependency patterns analyzed from the experiment results. Key words: color image segmentation, natural scenes, color spaces, k-means clustering
References
[1] Busin, L., Vandenbroucke, N., and Macaire, L., "Color spaces and image segmentation", Advances in Imaging and Electron Physics 151, 65–168 (2008). [2] Cheng, H., Jiang, X., Sun, Y., and Wang, J., "Color image segmentation: advances and prospects", Pattern Recognition 34(12), 2259–2281 (2001). [3] Busin, L., Shi, J., Vandenbroucke, N., and Macaire, L., "Color space selection for color image segmentation by spectral clustering", IEEE International Conference on Signal and Image Processing Applications 22(8), 262–267 (2009). [4] Fernando E. Correa-Tome, Raul E. Sanchez-Yanez and Victor Ayala-Ramirez, "Comparison of perceptual color spaces for natural image