彩色图像球空间中改进DCNON分割算法的研究

Improved DCNON Segmentation Algorithm Research in Color Image Sphere Coordinate

提出一种基于改进振荡神经网络的彩色图像分割方法。该方法将彩色图像从RGB的三维笛卡尔空间转换到球坐标空间,去除由光照和纹理引起的背景噪声,构成平滑的相位灰度图;将相位灰度图影射到动态耦合振荡神经网络的二维平面上;利用改进结构的算法对相位灰度图进行分割,得到稳定、快速和可靠的分割结果。

This paper introducd a method of advanced dynamically coupled neural oscillator network based on color image segmentation. This method first converses RGB of color image from three-dimensional Cartesian coordinates to spherical coordinates, aiming to remove background noise due to illumination and texture to generate smooth gray-level image of angle. Then, the advanced dynamically coupled neural oscillator network will do segmentation for the gray-level image of angle, to generate robust, fast and reliable segmentation.