景深合成算法研究及其在超景深显微镜设计中的应用
发布时间：2018-08-28 12:06
【摘要】：随着人类对微观世界的观察需求,显微镜被广泛应用,但是传统显微镜在使用过程中有诸多不便,如对焦缓慢、景深限制等。基于图像序列的超景深图像融合算法,是根据采集于同一场景但部分对焦清晰的图像序列,合成得到一副全局对焦清晰的超景深图像。这样的算法可以打破传统光学显微镜的景深限制,设计出方便实用的超景深显微镜。超景深显微镜在放大倍数较高的情况下能方便清晰地观察起伏较大的样品,在医疗、生物、工业生产等领域广泛应用。故本文研究了多种图像融合算法,如基于图像金字塔的图像融合算法、基于小波变换的图像融合算法和基于最优化的图像融合算法,并又提出一种高效的景深扩展算法用于超景深显微镜设计。首先,在本文提出的方法中,通过设计一个根据图像纹理和分辨率自适应的聚焦度量用于检测图像序列中所有图像的所有位置的聚焦清晰程度。为了增强系统鲁棒性,针对显微镜光学特点,把对焦度量结果用一个高斯分布表示,用以模拟还原真实场景的深度信息和对焦状况。然后,为了降低噪声和特殊样品对融合结果的影响,以便适应更多的应用场景,本文的方法将图像中每个像素根据不同的聚焦情况分为三种类型,并针对不同的聚焦特点使用不同的融合规则。这里称这个步骤为局部加权平均的融合规则。在图像区域清晰且噪声少的情况下仅仅使用一张图像执行融合;在图像区域含有噪声的情况下使用两张或者三张比较清晰的图像执行融合;在信息质量差的图像区域使用附近区域的规则替代当前位置的融合规则。最后,为了评估本文所提出算法的有效性,本文在大量模拟数据和真实数据上进行广泛的实验,最终的定量与定性结果表明本文提出的景深图像融合算法是简单且极其有效的。此外,大量的对比算法定量分析证明了本文所提出的算法无论是效果上还是速度上都远优于传统的方法。值得关注的是,一台利用本文提出的算法的超景深数字光学显微镜被实际设计出来并已经投入工业界使用,对实际工业样品的真实显示结果令客户满意,且获得了大额的经济效益。
[Abstract]:Microscopes are widely used with the need of human being to observe the microcosm, but the traditional microscopes have many inconveniences, such as slow focusing, limited depth of field and so on. The fusion algorithm of hyper-depth image based on image sequence is based on the image sequence which is collected in the same scene but partially focused, and a set of hyper-depth image with global focus is synthesized. This algorithm can break through the limit of depth of field of traditional optical microscope and design a convenient and practical hyper-depth microscope. The hyperfield depth microscope can be used in many fields such as medical treatment, biology, industrial production and so on, which can easily and clearly observe the large undulating samples under the condition of high magnification. So this paper studies many image fusion algorithms, such as image fusion algorithm based on image pyramid, image fusion algorithm based on wavelet transform and image fusion algorithm based on optimization. An efficient depth of field extension algorithm is also proposed for the design of hyper-depth microscope. Firstly, in the proposed method, an adaptive focusing metric based on image texture and resolution is designed to detect the focus clarity of all the images in the image sequence. In order to enhance the robustness of the system, aiming at the optical characteristics of the microscope, the focus measurement results are represented by a Gao Si distribution, which is used to simulate the depth information and the focusing state of the real scene. Then, in order to reduce the effect of noise and special samples on the fusion results, in order to adapt to more application scenarios, each pixel in the image is divided into three types according to different focusing conditions. According to different focusing characteristics, different fusion rules are used. This step is referred to here as the fusion rule of local weighted average. When the image area is clear and the noise is low, only one image is used to perform fusion, and two or three clear images are used to perform fusion when the image region contains noise. The fusion rule of the current location is replaced by the rule of the nearby region in the image area with poor information quality. Finally, in order to evaluate the effectiveness of the proposed algorithm, extensive experiments are carried out on a large number of simulated and real data. The final quantitative and qualitative results show that the proposed algorithm is simple and extremely effective. In addition, a large number of quantitative analysis results show that the proposed algorithm is far superior to the traditional method both in effect and speed. It is worth noting that a digital optical microscope using the algorithm proposed in this paper has been designed and put into use in industry. The actual display results of the actual industrial samples are satisfactory to the customers. And obtained the economic benefit of large amount.
【学位授予单位】：安徽大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41
本文编号：2209347
[Abstract]:Microscopes are widely used with the need of human being to observe the microcosm, but the traditional microscopes have many inconveniences, such as slow focusing, limited depth of field and so on. The fusion algorithm of hyper-depth image based on image sequence is based on the image sequence which is collected in the same scene but partially focused, and a set of hyper-depth image with global focus is synthesized. This algorithm can break through the limit of depth of field of traditional optical microscope and design a convenient and practical hyper-depth microscope. The hyperfield depth microscope can be used in many fields such as medical treatment, biology, industrial production and so on, which can easily and clearly observe the large undulating samples under the condition of high magnification. So this paper studies many image fusion algorithms, such as image fusion algorithm based on image pyramid, image fusion algorithm based on wavelet transform and image fusion algorithm based on optimization. An efficient depth of field extension algorithm is also proposed for the design of hyper-depth microscope. Firstly, in the proposed method, an adaptive focusing metric based on image texture and resolution is designed to detect the focus clarity of all the images in the image sequence. In order to enhance the robustness of the system, aiming at the optical characteristics of the microscope, the focus measurement results are represented by a Gao Si distribution, which is used to simulate the depth information and the focusing state of the real scene. Then, in order to reduce the effect of noise and special samples on the fusion results, in order to adapt to more application scenarios, each pixel in the image is divided into three types according to different focusing conditions. According to different focusing characteristics, different fusion rules are used. This step is referred to here as the fusion rule of local weighted average. When the image area is clear and the noise is low, only one image is used to perform fusion, and two or three clear images are used to perform fusion when the image region contains noise. The fusion rule of the current location is replaced by the rule of the nearby region in the image area with poor information quality. Finally, in order to evaluate the effectiveness of the proposed algorithm, extensive experiments are carried out on a large number of simulated and real data. The final quantitative and qualitative results show that the proposed algorithm is simple and extremely effective. In addition, a large number of quantitative analysis results show that the proposed algorithm is far superior to the traditional method both in effect and speed. It is worth noting that a digital optical microscope using the algorithm proposed in this paper has been designed and put into use in industry. The actual display results of the actual industrial samples are satisfactory to the customers. And obtained the economic benefit of large amount.
【学位授予单位】：安徽大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41
【参考文献】
相关期刊论文 前2条
1 王莉;蒋洪;孙丽丽;;显微镜的发展综述[J];科技信息;2009年11期
2 屈小波;闫敬文;谢国富;朱自谦;陈本刚;;A novel image fusion algorithm based on bandelet transform[J];Chinese Optics Letters;2007年10期
，本文编号：2209347
本文链接：https://www.wllwen.com/shoufeilunwen/xixikjs/2209347.html
