デジタルホログラフィック顕微鏡を用いた分解能向上及びノイズ低減による対象物の正確な3次元情報取得に関する画像処理手法

Le résumé fourni par la source

Holography is a representative technique that uses coherence to obtain the phase information of light. In this technique, reference and object beams are generated using coherent plane waves or spherical waves, and then the interference pattern between them is recorded on a photosensitive plate. Digital holography （DH） has basically the same principle as that of holography, but it uses image sensors instead of films for recording. DH has been widely applied in many applications such as 3D image encryption, 3D image recognition, digital holographic reconstruction, and digital holographic microscopy （DHM）. Among these applications, DHM is used for disease diagnosis, 3D profiling of microstructures, and microbial research because it can reconstruct the thickness profile of a micro-object with high-depth resolution. However, DHM has a problem in that it cannot obtain accurate 3D information due to noise. The cause of this noise can be divided into a hardware cause and a software cause. A representative hardware problem of DHM is noise by inaccurate focal plane. Since the depth of field （DOF） of the objective lens decreases as the magnification increases, it is difficult to accurately position the object on the focal plane, and due to this problem, accurate object information cannot be obtained. In addition, a software problem is the trade-off relationship between detailed information about an object and phase error. When a wide size of the windowed sideband is used, the high spatial frequency information of the object is included, but the phase error increases because DC spectrum information is also included. On the other hand, when a narrow window size is used, the phase error decreases, but the high spatial frequency information of the object is lost. Since DHM is used in fields where accuracy is important, such as a technology for diagnosing of disease, it is necessary to minimize noise and provide detailed and accurate 3D information. To solve these problems, I propose image processing methods to acquire accurate 3D information of objects with enhanced resolution and reduced noise for DHM in this thesis. I analyze the problems of DHM by hardware and software points of view, and propose methods to solve the problems. In this thesis, I propose a method to measure the refractive index of the surrounding medium of an object that needs to be accurately known for accurate height information in real-time using a specimen made with an object of known exact refractive index and height. Also, I propose a method of finding an accurate focal plane using noise caused by the depth of field of the objective lens. In addition, I propose algorithms for various image processing methods that solve the trade-off relationship between detailed information of an object and phase error using the image processing technique on how to use the windowed sideband. In addition, I explain how to make statistical analyses by automatically classifying the obtained 3D data.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.