首页|基于多尺度融合和分数阶微分的工地图像增强

基于多尺度融合和分数阶微分的工地图像增强

扫码查看
建筑工地采集的图像通常会有色偏、对比度低和纹理模糊等问题,从而导致无法获得良好的人眼视觉体验和正确的机器视觉处理结果.为此,提出一种基于改进的多尺度融合和自适应分数阶微分的工地图像增强算法.针对工地图像的纹理模糊特征对多尺度融合算法和自适应分数阶微分算法进行改进,采用全局和局部对比度增强图像替换两幅输入图进行多尺度融合,进一步提高图像的对比度;在HSV颜色空间下仅对V通道分量进行自适应分数阶微分且与原始图像进行加权融合,实现在不改变原本颜色的情况下进行纹理增强和弱化伪影现象.实验结果表明,本文算法增强后的图像拥有更自然的色调、更高的对比度和更强的细节表达能力,优于其他图像增强算法.由此,本文方法能够快速且高效地增强低质量的工地图像,提高后续机器视觉处理的精度和速度,在解决工地图像质量欠佳的问题中发挥重要作用.
Construction Site Image Enhancement Based on Multi-scale Fusion and Fractional Differential
Images from construction sites often have color bias,low contrast and blurred texture,making it impossible to obtain a good human visual experience and correct machine vision processing results.Therefore,an image enhancement algorithm based on multiscale fusion and adaptive fractional differentiation is proposed.In this paper,the multi-scale fusion algorithm and the adaptive fractional differential algorithm are improved according to the texture fuzzy feature of the site image,in the HSV color space,only the V-channel components are differentiated by adaptive fractional order and fused with the original image,it can enhance and weaken the artifacts without changing the original color.The experimental results show that the enhanced image has more natural hue,higher contrast and better detail expression ability,which is better than other image enhancement algorithms.The method proposed in this article can quickly and efficiently enhance low-quality construction site images,thereby improving the accuracy and speed of subsequent machine vision processing,and playing an important role in solving the problem of poor construction site image quality.

image enhancementmulti-scale fusionadaptive fractional differentialsite image

林咸磊、陈国栋、佘明磊、牟宏霖、林进浔

展开 >

福州大学 物理与信息工程学院,福建 福州 350108

南平武沙高速公路有限责任公司,福建 南平 353000

福建数博讯信息科技有限公司,福建 福州 350002

图像增强 多尺度融合 自适应分数阶微分 工地图像

福建省科技计划引导性项目福建省科技型中小企业创新资金项目

2021H00132021C0019

2024

贵州大学学报(自然科学版)
贵州大学

贵州大学学报(自然科学版)

CSTPCD
影响因子:0.396
ISSN:1000-5269
年,卷(期):2024.41(4)