针对光照变化和阴影对图像分割的不利影响问题,提出了一种基于矢量量化和区域生长的彩色图像分割新算法。该算法不仅考虑了彩色图像的颜色信息,而且也考虑了彩色图像的空间信息。该算法首先利用一种修改的GLA算法对彩色图像进行量化,并根据彩色图像量化的结果选取种子像素;然后基于矢量角相似性准则,并结合像素空间邻接信息,对每一个种子像素进行区域生长;最后利用模糊C-M eans算法来对未能归类的剩余像素进行分类。实验表明,该算法不仅可以在很大程度上克服光照变化及阴影对图像分割的不利影响,而且分割结果与人的主观视觉感知具有良好的一致性。

In order to solve the problem of the negative effects of shading and uneven illuminations in the color image on segmentation,this paper proposes a novel algorithm for color image segmentation based on vector quantization and region growing.The algorithm takes into account both color information and spatial information of color images,which avoids the insufficiency of previous methods which have considered either the color information of the image,or the spatial information of the image.Firstly,the color image is quantized by using a modified generalized Lloyd algorithm(GLA),and then selects seed pixels in terms of the quantization results of color images.Secondly,from each seed pixel region growing is performed based on the vector angle similarity criterion.Finally,fuzzy C-means(FCM) algorithm is used to classify the remaining pixels which have not been classified into regions being formed earlier.The experiments show that the negative effects of shading and uneven illuminations in the color image on segmentation can be overcome to a great extent by using the proposed algorithm.Color image segmentation results of the proposed approach hold favorable consistency in terms of human visual perception.