摘要
针对偏暗、低对比度图像视觉显著性图提取效果不理想的问题,提出一种基于改进的直方图均衡化的图像显著性图提取算法.作者通过分析发现某些偏暗、低对比度图像的视觉显著性图提取效果不理想的主要原因是由于图像背景与前景差异不明显导致的,因此本文提出采用直方图均衡化方法对图像的质量加以改善,并对传统直方图均衡化算法存在的曝光过度等问题有针对性的进行了改进,并以此提高图像前景与背景之间的差异,使得图像的显著区域更加突出,再对处理后的图像进行显著性图的提取.为了验证算法的有效性,本文在ASD1000数据集和LBE数据集上分别进行了仿真实验,实验结果表明本文提出的算法是有效的,并具有较高的鲁棒性和准确性.
In order to solve the saliency detection is not ideal for dark and low contrast image under the natural scene. So,this paper proposing a new saliency detection algorithm based on the improved histogram equalization. The author found that dark and low contrast image＇s saliency recognition is not well. Because these image＇s background and foreground have little difference. Applying the histogram equalization to improve the brightness and contrast,at the same time,optimizing the overexposed and other problems of the traditional histogram equalization method,increasing the difference between background and foreground,prompting the saliency region more prominent,and extracting the processed image,the results have a greatly improved. So as to verifying the validity of the algorithm,we conducted the experiment on the ASD1000 data set and LBE that is small data sets by laboratory established,respectively,the experimental results show that this method has high robustness and accuracy.
作者
罗强
吴俊峰
于红
孙建伟
张美玲
LUO Qiang;WU Jun-feng;YU Hong;SUN Jian-wei;ZHANG Mei-ling(College of Information Engineedng, Dalian Ocean University, Dalian 116023, China;Key Laboratory of Marine Information Technology in Liaoning Province,Dalian 116023, China)
出处
《小型微型计算机系统》
CSCD
北大核心
2018年第5期1092-1096,共5页
Journal of Chinese Computer Systems
基金
海洋渔业大数据管理与集成关键技术研究项目(2015A11GX022)资助
关键词
直方图均衡化
对比度增强
显著性提取
histogram equalization
contrast enhancement
saliency detection
