A saliency detection method for a 360-degree image

The invention relates to a saliency detection model method for a 360-degree image, which is characterized in that the 360-degree image is divided into super pixel blocks by using a simple linear iterative clustering SLIC algorithm, and then a color space is converted into a CIE Lab color space for extracting brightness, color and texture sensing features; then, based on Gestalt theory, two components of saliency detection are calculated, which are feature contrast and boundary connectivity; and finally, the 360-degree saliency map is obtained by fusing the feature contrast map and the boundaryconnectivity map. Experiments on the published 360-degree image dataset show that the proposed model can predict the salient region of the image more accurately. 本发明涉及种对于360度图像的显著性检测模型方法，其特征在于：利用简单线性迭代聚类SLIC算法将360度图像分割成超像素块，然后将颜色空间转化为CIE Lab颜色空间，用来提取亮度，颜色和纹理感知特征；接着基于格式塔理论，计算显著性检测的两个组成部分，即特征对比度和边界连接度；最后通过融合特征对比度图和边界连接度图求得360度显著图。在公开的360度图像数据集上的实验表明，该提出的模型能够较为准确地预测图像的显著区域。