基于双阈值canny均衡化算法的太赫兹图像增强

存在于微波与远红外之间的太赫兹波，因其无损害，稳定性高等独特性质使太赫兹光谱与成像技术在近几年来得到了迅猛的发展。太赫兹波独有的无损伤检测特性，在安全检测方面具有良好的发展前景，获得广大学者的研究和关注。经过太赫兹成像系统获得的太赫兹图像，虽然可以识别出隐藏的武器或其他金属制品，但是太赫兹图像的对比度和清晰度均较差，不能完全符合人眼的视觉效果，也不利于机器识别。目前，对太赫兹图像质量的提高和改善，成为太赫兹成像技术长远发展和广泛应用的关键问题。实验采用太赫兹波投射式成像系统对藏于物体中的金属心型吊坠和金属箭头进行成像，扫描步长0.5 mm，由于太赫兹光源大，能量起伏等系统缺陷，以及外部环境的复杂与干扰，导致成像所得图像均有背景噪声严重，边界模糊等问题，成像质量较差。提出一种基于双阈值的canny均衡化太赫兹图像增强算法，根据太赫兹图像自身性质限制，确定阈值和对图像均衡化的范围，实现图像降噪并引入双阈值canny算法和梯度幅值算法，使图像的对比度和清晰得到整体提高，并保留和突出太赫兹图像的细节信息，获得高分辨率、边缘清晰的图像。实验表明该算法对太赫兹图像具有良好的降噪效果，能够保留图像细节信息，图像对比度和图像质量得到增强和提高，同时增强了太赫兹成像技术对隐藏缺陷或隐藏物体的辨别能力和透视能力，为其在安检应用方面提供必要保证。

Based on Double Threshold Canny Equalization Algorithm for Terahertz Image Enhancement

Terahertz wave (THz), which exists between microwave and far infrared, has made the terahertz spectroscopy and imaging technology develop rapidly in recent years because of its non-damage property and high stability. The unique feature of non-invasive detection of terahertz wave has a perfect prospect in the field of security detection, and has attracted the attention of scholars. Although THz images obtained by terahertz imaging system can identify hidden weapons or other metal products, the THz image contrast and clarity are poor, which can not completely accord with human visual effect, and is not conducive to machine recognition. At present, the enhancement and improvement of terahertz image quality are the key to the long-term development and wide application of THz imaging technology. By using terahertz projection imaging system with scanning step 0.5mm, we imagine the objects of the metal pendant and metal arrow hidden in clothes at the same experimental conditions. Because of the system defects such as terahertz light source, energy fluctuation as well as the complexity and interference of the external environment, the imaging of the imaging system we get has serious background noise and blurred boundary, and the imaging quality is poor. This paper presents an algorithm of Canny equalization based on double threshold for THz image enhancement. According to the property limitations of terahertz image itself, this algorithm determines the relevant thresholds and the range of image equalization to realize the image denoising. The algorithm introduces the double threshold Canny algorithm and the gradient amplitude algorithm to improve the contrast and clarity of the image, preserve and highlight the detail information of the terahertz image, as well as obtain the high resolution and clear edge image. The experimental results show that we successfully obtain the terahertz images with improved definition, lower noise and fully detailed information from tow low-resolution (LR) terahertz images. The quality of images is improved.This work also enhances the discrimination ability and perspective capability for the hidden defects or hidden objects in the terahertz images, which demonstrates the enormous potential that the terahertz imaging provides the application in the security check.