In this paper a rapid focus detection technique is developed for objects imaged using digital in-line holograms. It differs from previous approaches in that it is based directly on the spectral content of the object images and does not need a full reconstruction of the actual images. It is based on new focus metrics defined as the l 1 norms of the object spectral components associated with the real and imaginary parts of the reconstruction kernel. Furthermore, these l 1 norms can be computed efficiently in the frequency domain using a polar coordinate system, yielding a drastic speedup of about two orders of magnitude compared with image-based focus detection. The subsequent reconstruction, when done selectively over these detected focus distances, leads to significant computational savings. Focus detection results from holograms of plankton are demonstrated showing that the technique is both accurate and robust.

Sampling the vast volumes of the ocean requires tools capable of observing from a distance while retaining detail necessary for biology and ecology, ideal for optical methods. Algorithms that work with existing SeaBED AUV imagery are developed, including habitat classi…cation with bag-of-words models and multi-stage boosting for rock…sh detection. Methods for extracting images of …sh from videos of longline operations are demonstrated.A prototype digital holographic imaging device is designed and tested for quantitative in situ microscale imaging. Theory to support the device is developed, including particle noise and the e¤ects of motion. A Wigner-domain model provides optimal settings and optical limits for spherical and planar holographic references.Algorithms to extract the information from real-world digital holograms are created. Focus metrics are discussed, including a novel focus detector using local Zernike moments. Two methods for estimating lateral positions of objects in holograms without reconstruction are presented by extending a summation kernel to spherical references and using a local frequency signature from a Riesz transform. A new metric for quickly estimating object depths without reconstruction is proposed and tested. An example application, quantifying oil droplet size distributions in an underwater plume, demonstrates the e¢ cacy of the prototype and algorithms. Course work at MIT and WHOI was helpful throughout this thesis. In particular, course materials and projects from Dr. Antonio Torralba (computer vision, detection, and boosting), Dr. Frédo Durand (computational imaging, segmentation, and probability), Dr. Rosalind Picard (machine learning, pattern recognition, and probability), Dr. Alan 4 Edelman (parallel computing and algorithms), and Dr. Barbastathis (optical systems) are re ‡ected directly, in several cases extended after the course to become complete sections of this thesis.The MIT Museum has provided unique outreach opportunities for the ideas presented in this thesis. Seth Riskin originally involved our lab in holography activities at the MIT Museum, then Kurt Hasselbalch later proposed and guided the development of an interactive display on plankton holography. The computational approaches I learned for the museum display led to further scienti…c development and spurred the use of GPUs for processing; Section 3.3.5 is a direct consequence, as is the majority of the data processing performed throughout this thesis.In reference to data processing, Matlab/MEX and NVIDIA's CUDA have been daily cornerstones for which I am continually thankful. Finally, mad props go to Dr. José "Pepe" Domínguez-Caballero, the most in ‡uential 5 person in my scienti…c development while at MIT. Pepe taught me digital holography and worked one-on-one with me throughout the beginning of my grad student career. He continued to be involved as a fellow holographer, a mentor, and a friend, discussing ideas, encouraging active experimentation, maintaining a curiosity about optics, and o¤ering eno...

No abstract