低对比度全景球面图像目标分割方法研究

受到光照、设备等外界条件的影响,得到的全景球面图像对比度通常较低,当前目标分割方法无法解决外界环境的干扰问题,导致分割结果精度低,分割效果不佳。为此,提出一种新的低对比度全景球面图像目标分割方法,通过PCNN模型对低对比度全景球面图像进行对比度增强处理,依据人眼视觉特征,通过对数变换映射函数把全景球面图像的亮度调整至一个合适的视觉范围内。介绍了均值偏移法的理论基础,通过对特征空间中样本点的聚类,获取模式点。通过均值偏移法将空间上相邻和色彩相同的像素划分至一类,找到不同颜色的聚类点,从而实现低对比度全景球面图像目标分割。实验结果表明,采用所提方法对低对比度全景球面图像进行分割,不仅分割效果好,而且分割精度高。

Low contrast panoramic spherical target image segmentation method research

under the influence of lighting, equipment and other external conditions, get a panoramic view of spherical image contrast is low, usually the current target segmentation method can"t solve the problem of the interference of the external environment, lead to low accuracy segmentation result, split the effect not beautiful.For this, put forward a new kind of low contrast panoramic spherical target image segmentation method, through the PCNN model to simulate the low contrast panoramic spherical image contrast enhancement processing, according to human visual characteristics, through the logarithmic transformation mapping function to the panorama of spherical image brightness adjustment to an appropriate visual range.Introduced the basic theory of the mean shift method, through to the sample points in the feature space clustering, access mode.By the mean shift method on the space adjacent and the pixel is classified to the same kind of color, find a different color clustering points, thus achieving low contrast panoramic spherical image segmentation.The experimental results show that the proposed method of low contrast panoramic spherical image segmentation, segmentation effect is good, not only and segmentation precision is high.