多特征融合的医学影像颜色感知算法

摘要

Due to the facts that the existing medical images are mainly represented in the form of gray scale images,with the monotonous feature,and cannot express the image information adequately,a color perception algorithm of medical images based on multi-features fusion was proposed.First,by extracting the gradient features of the gray images in multiple directions and fusing original brightness feature,more information about the images was retained.Next,a hierarchical structure was constructed to reduce the computation cost by selecting representative pixels in similar region.Then,the low-dimensional coordinates for all pixels on the top layer were calculated by using manifold-based technique and the interpolation computation was conducted from top to bottom.Finally,low-dimensional coordinates were projected into the color space to gain the corresponding color medical images.The experimental results on normal MRI images,normal CT images and MRI images with disease,demonstrated that the obtained color medical images have rich information by our algorithm.Moreover,compared with the traditional color transfer algorithm,the color medical images by using our algorithm has higher clarity and better target background contrast indicators.%针对现有的医学影像主要以灰度图像的形式呈现,特征单一,不能充分表达图像信息等问题,提出多特征融合的医学影像颜色感知算法.首先通过提取多个方向的梯度特征以及融合原始亮度特征保留图像更多信息,在相似区域选择代表点,建立像素点的层次结构减少计算量;然后对顶层像素点利用流形学习和自顶而下的插值方法计算所有像素点的低维坐标;最后将低维坐标投影到颜色空间得到相应的彩色医学图像.在正常MRI影像、正常的CT影像以及带有疾病的MRI影像上进行实验的结果表明,文中算法对医学灰度图像进行着色后具有丰富的色彩信息;相比传统的颜色迁移算法,该算法具有更高的清晰度和目标背景对比度指标.