低对比度全景球面图像目标分割方法

Low Contrast Panoramic Spherical Target Image Segmentation Method

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。

U nder the influence of lighting, equipment and other external conditions, get a panoramic view ofspherical image contrast is l o w, usually the current target segmentation mettiod can＇ t solve the problem of theinterference of the external environment, lead to low accuracy segmentation result, split the effect this, a new kind of low contrast panoramic spherical target image segmentation metiiod was put forwarPCNN model to simulate the low contrast panoramic spherical image contrast enhancement processing, acchuman visual characteristics, through the logarit!imic transformation map p i n g function to the image brightness adjustment to an appropriate visual range. Introduced the basic theory of the m e a n shift m e t h o d,through to the sample points in the feature space clustering, access m o d e. B y the m e a n shift method on the spaceadjacent and the pixel is classified to the s a m e kind of color, find a differelow contrast panoramic spherical image segmentation. T h e experimental results s how that t!ie proposcontrast panoramic spherical image segmentation, segmentation effect is g o o d, not only and segmeniMhigh.