摘要
针对目前基于模式噪声方法处理夜间场景下图像来源检测准确效果较差问题,提出一种综合纹理和亮度的夜间场景图像来源检测方法。首先基于夜间场景图片会存在同张图片不同区域会有不同纹理和亮度从而会影响提取模式噪声质量不同这一理论依据,将图片根据纹理和亮度分成若干大小相等的块;然后对不同块按纹理平缓亮度良好、纹理复杂亮度差、纹理平缓亮度差、温度复杂亮度良好四种情况处理,其中对于纹理复杂的区域提取的噪声要抑制纹理干扰、对于亮度条件差的区域提取的模式噪声需要增强、对于亮度差且纹理又复杂的区域提取的模式噪声、既要抑制纹理的干扰又要增强模式噪声。实验结果表明,本文方法在夜间场景情况对图像检测整体识别率均在在80%以上,与传统模式噪声提取算法相比,该算法正确率能够提高4到12个百分点。
As the method based on pattern noise cannot deal with the image source detection in night effectively,this paper proposes a method of detecting the source of night scene images with comprehensive texture and brightness.First,based on the nocturnal scene,there will be different textures and lumi- nances in different regions of the same picture,which will affect the difference of the quality of the extraction pattern.The images are divided into pieces with equal size according to the texture and brightness.Four cases are treated respectively,in which the noise extracted from complex texture regions should be suppressed by texture interference,the mode noise extracted from regions with poor luminance conditions needs to be enhanced,for the mode noise extracted from regions with low brightness and complex texture,the interference of texture should be suppressed,while the mode noise should be enhanced at the same time.The experimental results show that the overall recognition rate of this method is above 80%in the night scene.Compared with the traditional mode noise extraction algorithm,the accuracy of the algorithm can be enhanced by 4%--12%.
作者
王瑞昆
柯永振
陈凌翔
WANG Rui-kun;KE Yong-zhen;CHEN Ling-xiang(School of Computer Science and Software Engineering,Tianjin Polytechnic University,Tianjin 300387,China)
出处
《光电子．激光》
EI
CAS
CSCD
北大核心
2018年第12期1358-1364,共7页
Journal of Optoelectronics·Laser
基金
国家自然科学基金(61602344)资助项目
关键词
图片来源检测
夜间场景
模式噪声
纹理
亮度
image source detection
night scene
pattern noise
texture
brightness
