【作者】 刘鹏；

【导师】 白丽平；

【作者基本信息】 广东工业大学 ， 机械工程， 2015， 硕士

【摘要】 静脉是人体传输养分和新陈代谢的重要参与者。体内与体外皮下的静脉分布能够表征人体的健康状况,例如现代医学证明了由于癌变区域需要大量养分输送故该区域的静脉分布异常密集；对于一些肥胖的婴儿,清晰的静脉分布有助于医生进行静脉注射。血红红细胞具有对近红外光的强吸收特性。传统的静脉显像方法使用近红外成像进行静脉显像,但该类方法图像获取的设备大、无法应用于体内并且在凸显静脉的前提下丢失了非静脉区域表皮的原色,本文的目的是研究一种能够应用于体内的静脉图像获取方法,并提出一种既能保持表皮原色又能凸显静脉的算法。要达到上述目的,首先需要提出一种能够适用于体内和体外的装置,利用该装置同时获取近红外与可见光图像；在获取到图像之后,对两幅图像进行配准。最后使用图像融合的方法将配准后的近红外和可见光图像进行融合。为了同时获取近红外图像与可见光图像,需要将近红外相机与普通的彩色相机构成双目系统。但这种方法无法做到超小型集成化,因此不能与内窥镜系统相结合应用于体内静脉显像。本文提出一种基于多光谱成像的微型静脉显像技术,该方法使用基于MEMS技术的多光谱图像获取装置来同时获取近红外与可见光图像。在图像融合之前必须进行图像配准。由于两幅图像的光谱特性及其成像装置有差异,会由位置不同导致两幅图像存在明显的视差,它们是明显的异源图像。此外,图像间可能还存在几何形变、噪声干扰等因素。需要对两幅图像进行预处理和图像配准。基于Speeded Up Robust Features (SURF)描述子具有尺度不变性和旋转不变性,它们能够很好地适用于异源图像的配准,所以本文使用该算法来完成近红外与可见光图像的配准。最后,本文提出一种保持原色的近红外与可见光图像融合算法,该算法主要使用直方图映射将近红外与彩色图像进行映射以保证亮度的一致性,然后使用Principal component analysis(PCA)色彩残差增强来保证融合图像与原可见光图像的色彩均一,快速双边滤波算法进行可见光与近红外图像细节保持,随后将处理后的结果进行结合得到既能保持原色又能凸显静脉的融合结果。更多还原

【Abstract】 Veins are important participants of human nutrition transmitting and metabolism. Inside body and subcutaneous vein distribution can reflect the human’s state of health, such as modern medicine proved that cancerous regions need a lot of nutrients so the veins distribute anomaly and intensively in this regions. For some obese baby, clear vein distribution could help doctor to inject intravenous. The red blood cells can absorb near infrared strongly. Traditional vein imaging method adopted near infrared imaging to get venous imaging. But image equipment of this method is too big to apply it to detect inside body. Moreover, it will lose the primary colors of the non-intravenous regional epidermis in the premise of prominent vein. The purpose of this paper was to get an image acquisition method that can be used in the body of vein, and proposed an algorithm which can not only keep the skin color but also highlights vein.To achieve the method and algorithm above, First, this paper needs a suitable device that is used both inside and outside the person’s body. This device was used to acquire near infrared and visible light image at the same time; then, registers the two images. Finally, adopts image fusion method to fuse the registers images containing near infrared image and visible light image.In order to obtain near infrared image and visible light image simultaneously, people needed to take near infrared camera and ordinary color camera to constitute a binocular system. But this method could not obtain an ultra-small integrated device, it also cannot be combined with the endoscope system to get inside body images. This paper proposed a kind of miniature vein imaging method based on a multi spectral imaging technology. The multi-spectral imaging device can got both near infrared and visible light images through MEMS technology.Before fusing the images, the two images should be registries before images fusion. There were differences between the two pieces of image spectral characteristics and imaging device. Moreover, the different positions existed obvious parallax in two images, which were obviously heterologous images. Besides, there may be geometric distortion, noise and other interference factors. Thus, it was needed to pretreat and register of the two images. The SURF feature descriptor has the character of scale invariance and rotation invariance. They could be well used in multi-sensor image registration, so this paper used this algorithm to register the near infrared and visible light images.At last, this paper proposed a near infrared and visible image fusion algorithm maintained color. This algorithm mainly used histogram mapping of near infrared and color image mapping to ensure the consistency of brightness, then use the PCA color residual enhance to ensure the fusion images kept original color with the visible light images. Fast bilateral filtering algorithm kept the details between visible light and near infrared image. Subsequently, the processed results are combined to obtain the fusion image of maintaining color and highlight the vein at the same time.更多还原