基于改进神经网络的弱光图像自动分割算法
Automatic Segmentation Algorithm for Low-Light Images Based on Improved Neural Networks
李胜 1呼家龙 1蒋国庆1
作者信息
- 1. 马鞍山职业技术学院 电子信息系,安徽 马鞍山 243031
- 折叠
摘要
弱光条件下图像的对比度低、细节模糊、噪声干扰大.为了使分割后的弱光图像边缘更加整齐,分割目标更加完整,提出一种基于改进神经网络的弱光图像自动分割算法.采用SLIC方法展开弱光图像超像素,根据弱光图像中所存在的像素相似性,对所生成的全部超像素块展开超像素块融合处理;并结合改进的神经网络,分割出弱光图像的目标图像与背景图像,实现弱光图像自动分割.实验结果表明,所提方法的弱光图像超像素融合效果好、分割精度高.
Abstract
Under low-light conditions,images suffer from low contrast,blurred details,and substantial noise interference.To achieve neater edges and more complete segmentation targets in low-light images,an automatic segmentation algorithm for low-light images based on an improved neural network is proposed.The SLIC method is employed to perform superpixel segmentation on low-light images.Subsequently,all generated superpixel blocks are fused based on the pixel similarity within the low-light images.Furthermore,by incorporating an im-proved neural network,the target image and background image of the low-light image are segmented,enabling automatic segmentation of the low-light image.The experimental results demonstrate that the proposed method ex-hibits a good superpixel fusion effect for low-light images,achieves high segmentation accuracy,and is suitable for automatic segmentation tasks under low light conditions.
关键词
弱光图像/图像分割/改进神经网络/SLIC方法/超像素分割Key words
low-light image/image segmentation/improved neural networks/SLIC method/hyperpixel segmentation引用本文复制引用
出版年
2024
NETL
NSTL
万方数据