基于凸包计算和颜色特征的显著性检测算法

An saliency detection algorithm via convex hull calculation and color features

针对传统的显著性检测算法存在区域亮度不够、显著区域不准确、有背景噪声等问题,提出一种基于凸包计算和颜色特征的显著性检测算法.考虑到图像在不同的颜色空间中颜色取值范围不同,首先,在多种颜色空间经过超像素分割得到区域对比图;然后,在CIELAB颜色空间中得到平滑的通道差值图;其次,利用颜色增强的Harris形成凸包得到中心先验图以及凸包结构图;最后,将4种显著图特征融合并优化得到最终显著图.新算法更加接近人工标注图的结果,不仅能够从背景中分离出显著物体、抑制背景干扰和凸出显著区域,而且获得全分辨率的显著图.在公开的图像数据集上将新算法与其他现有8种显著性检测算法进行对比,实验结果表明,新算法优于其他算法.

Aiming at the problems of traditional saliency detection algorithm,such as insufficient luminance,inaccurate saliency and background noise,we propose a saliency detection algorithm via convex hull calculation and color features.Considering that an image has different color value ranges in different color spaces,we firstly obtain a region contrast map by super-pixel segmentation in multi-color space.Then,a smooth channel difference map is obtained in the CIELAB space.Next,the color boosted Harris method is used to form a convex hull to obtain a center prior map and a convex hull structure map.The final saliency map is obtained by fusing four saliency map and optimizing them.This algorithm is more similar to the result of the artificial marked graph,it can not only separate the salient object from the background,restrain the background interference and highlight the salient region,but also obtain an full-resolution of the salient map.The proposed algorithm is compared with other 8 existing saliency detection algorithms on the public image dataset.Experimental results show that the algorithm is better than other algorithms.