红外技术2017，Vol.39Issue(5)：436-443,8.
综合改进Retinex和超像素分割的显著性检测算法
A Saliency Detection Algorithm Based on Improved Retinex and Super Pixel Segmentation
摘要
Abstract
In order to solve the existing problems of saliency map by significant detection algorithm, such as background clutter, noise and losing details, we put forward a comprehensive improvement based on Retinex algorithm and super pixel segmentation saliency detection algorithm. Firstly, the proposed algorithm uses the improved Retinex algorithm to pre-process the target image and obtain the enhanced image;Then, the proposed algorithm performs super-pixel segmentation on the processed image, performs wavelet transform on the super-pixel segmentation result, and generates the high-low frequency characteristic map of the original image, the high and low frequency feature map is filtered by the appropriate bilateral filter to generate the corresponding saliency map. Finally, the saliency map of the final target image is obtained by weighted high and low frequency saliency map. The experimental results show that the proposed method has obvious boundary and clear details, and it is not only effective to avoid the background noise interference, but also has better accuracy and recall than the other significant detection algorithms.关键词
显著图/显著性检测/Retinex/超像素分割/小波变换/双边滤波Key words
saliency map/saliency detection/Retinex/super segmentation/wavelet transform/bilateral filtering分类
信息技术与安全科学引用本文复制引用
叶潮流..综合改进Retinex和超像素分割的显著性检测算法[J].红外技术,2017,39(5):436-443,8.基金项目
安徽省高等学校自然科学研究一般项目(KJ2015B1105918). （KJ2015B1105918）
