Homography-guided stereo matching for wide-baseline image interpolation

Image interpolation has a wide range of applications such as frame rate-up conversion and free viewpoint TV. Despite significant progresses, it remains an open challenge especially for image pairs with large displacements. In this paper, we first propose a novel optimization algorithm for motion estimation, which combines the advantages of both global optimization and a local parametric transformation model. We perform optimization over dynamic label sets, which are modified after each iteration using the prior of piecewise consistency to avoid local minima. Then we apply it to an image interpolation framework including occlusion handling and intermediate image interpolation. We validate the performance of our algorithm experimentally, and show that our approach achieves state-of-the-art performance.

Image-to-Image Translation for Simplified MRI Muscle Segmentation

Deep neural networks recently showed high performance and gained popularity in the field of radiology. However, the fact that large amounts of labeled data are required for training these architectures inhibits practical applications. We take advantage of an unpaired image-to-image translation approach in combination with a novel domain specific loss formulation to create an “easier-to-segment” intermediate image representation without requiring any label data. The requirement here is that the task can be translated from a hard to a related but simplified task for which unlabeled data are available. In the experimental evaluation, we investigate fully automated approaches for segmentation of pathological muscle tissue in T1-weighted magnetic resonance (MR) images of human thighs. The results show clearly improved performance in case of supervised segmentation techniques. Even more impressively, we obtain similar results with a basic completely unsupervised segmentation approach.

Improving soft tissue imaging with volume-of-interest cone-beam CT

Current cone-beam CT systems acquire full field-of-view projections in which x-ray scatter degrades the contrast of soft-tissue in the reconstructed images. The objective of this work was to simulate volume-of-interest (VOI) imaging, which reduces scatter and dose to the patient through beam collimation, to investigate the improvements in soft-tissue visibility on the Gamma Knife Icon. The results indicated that as field size decreased, contrast and noise increased, leading to only modest improvements in the contrast-to-noise ratio when using the same initial photon fluence. A reconstruction framework called the interior virtual method was adapted to suppress truncation-induced artifacts and noise in the VOI image. In this framework the projection data were extrapolated using a cosine function, an intermediate image was reconstructed analytically, and virtual projections of the intermediate image were created for iterative reconstruction. The framework supports high quality VOI reconstruction and can allow clinicians to optimize dose for soft-tissue visualization.

Improving soft tissue imaging with volume-of-interest cone-beam CT

Current cone-beam CT systems acquire full field-of-view projections in which x-ray scatter degrades the contrast of soft-tissue in the reconstructed images. The objective of this work was to simulate volume-of-interest (VOI) imaging, which reduces scatter and dose to the patient through beam collimation, to investigate the improvements in soft-tissue visibility on the Gamma Knife Icon. The results indicated that as field size decreased, contrast and noise increased, leading to only modest improvements in the contrast-to-noise ratio when using the same initial photon fluence. A reconstruction framework called the interior virtual method was adapted to suppress truncation-induced artifacts and noise in the VOI image. In this framework the projection data were extrapolated using a cosine function, an intermediate image was reconstructed analytically, and virtual projections of the intermediate image were created for iterative reconstruction. The framework supports high quality VOI reconstruction and can allow clinicians to optimize dose for soft-tissue visualization.

OCR Text Extraction

This research tries to find out a methodology through which any data from the daily-use printed bills and invoices can be extracted. The data from these bills or invoices can be used extensively later on – such as machine learning or statistical analysis. This research focuses on extraction of final bill-amount, itinerary, date and similar data from bills and invoices as they encapsulate an ample amount of information about the users purchases, likes or dislikes etc. Optical Character Recognition (OCR) technology is a system that provides a full alphanumeric recognition of printed or handwritten characters from images. Initially, OpenCV has been used to detect the bill or invoice from the image and filter out the unnecessary noise from the image. Then intermediate image is passed for further processing using Tesseract OCR engine, which is an optical character recognition engine. Tesseract intends to apply Text Segmentation in order to extract written text in various fonts and languages. Our methodology proves to be highly accurate while tested on a variety of input images of bills and invoices.

Curvilinear features are important for animate/inanimate categorization in macaques

The current experiment investigated the extent to which perceptual categorization of animacy, i.e. the ability to discriminate animate and inanimate objects, is facilitated by image-based features that distinguish the two object categories. We show that, with nominal training, naïve macaques could classify a trial-unique set of 1000 novel images with high accuracy. To test whether image-based features that naturally differ between animate and inanimate objects, such as curvilinear and rectilinear information, contribute to the monkeys’ accuracy, we created synthetic images using an algorithm that distorted the global shape of the original animate/inanimate images while maintaining their intermediate features (Portilla and Simoncelli, 2000). Performance on the synthesized images was significantly above chance and was predicted by the amount of curvilinear information in the images. Our results demonstrate that, without training, macaques can use an intermediate image feature, curvilinearity, to facilitate their categorization of animate and inanimate objects.

Wide field-of-view volumetric imaging by a mesoscopic scanning oblique plane microscopy with switchable objective lens

Conventional light sheet fluorescence microscopy (LSFM), or selective plane illumination microscopy (SPIM), enables high resolution 3D imaging over a large volume by using two orthogonally aligned objective lenses to decouple excitation and emission. The recent development of oblique plane microscopy (OPM) simplifies LSFM design with only one single objective lens, by using off-axis excitation and remote focusing. However, most reports on OPM has a limited microscopic field of view (FOV), typically within 1×1 mm2. Our goal is to overcome the limitation with a new variant of OPM to achieve mesoscopic FOV. We implemented an optical design of mesoscopic scanning OPM to allow using low numerical aperture (NA) objective lens. The angle of the intermediate image before the remote focusing system was increased by a demagnification under Scheimpflug condition such that the light collecting efficiency in the remote focusing system was significantly improved. We characterized the 3D resolutions and FOV by imaging fluorescence microspheres, and demonstrated the volumetric imaging on intact whole zebrafish larvae, mouse cortex, and multiple Caenorhabditis elegans (C. elegans). We demonstrate a mesoscopic FOV up to ~6× 5×0.6 mm3 volumetric imaging, the largest reported FOV by OPM so far. The angle of the intermediate image plane is independent of the magnification. As a result, the system is highly versatile, allowing simple switching between different objective lenses with low (10x, NA 0.3) and median NA (20x, NA 0.5). Detailed microvasculature in zebrafish larvae, mouse cortex, and neurons in C. elegans are clearly visualized in 3D. The proposed mesoscopic scanning OPM allows using low NA objective such that centimeter-level FOV volumetric imaging can be achieved. With the extended FOV, simple sample mounting protocol, and the versatility of changeable FOVs/resolutions, our system will be ready for the varieties of applications requiring in vivo volumetric imaging over large length scales.