Sundaresh Ram

Bio: Sundaresh Ram is an academic researcher from University of Michigan. The author has contributed to research in topics: Image segmentation & Segmentation. The author has an hindex of 12, co-authored 39 publications receiving 349 citations. Previous affiliations of Sundaresh Ram include Velammal Engineering College & Cornell University.

Image Inpainting Using Nonlocal Texture Matching and Nonlinear Filtering

Abstract: Nonlocal texture similarity and local intensity smoothness are both essential for solving most image inpainting problems. In this paper, we propose a novel image inpainting algorithm that is capable of reproducing the underlying textural details using a nonlocal texture measure and also smoothing pixel intensity seamlessly in order to achieve natural-looking inpainted images. For matching texture, we propose a Gaussian-weighted nonlocal texture similarity measure to obtain multiple candidate patches for each target patch. To compute the pixel intensity, we apply the $\alpha $ -trimmed mean filter to the candidate patches to inpaint the target patch pixel-by-pixel. The proposed algorithm is compared with four current image inpainting algorithms under different scenarios, including object removal, texture synthesis, and error concealment. Experimental results show that the proposed algorithm outperforms the existing algorithms when inpainting large missing regions in images with texture and geometric structures.

Drive-Net: Convolutional Network for Driver Distraction Detection

Abstract: To help prevent motor vehicle accidents, there has been significant interest in finding an automated method to recognize signs of driver distraction, such as talking to passengers, fixing hair and makeup, eating and drinking, and using a mobile phone. In this paper, we present an automated supervised learning method called Drive-Net for driver distraction detection. Drive-Net uses a combination of a convolutional neural network (CNN) and a random decision forest for classifying images of a driver. We compare the performance of our proposed Drive-Net to two other popular machine-learning approaches: a recurrent neural network (RNN), and a multi-layer perceptron (MLP). We test the methods on a publicly available database of images acquired under a controlled environment containing about 22425 images manually annotated by an expert. Results show that Drive-Net achieves a detection accuracy of 95%, which is 2% more than the best results obtained on the same database using other methods.

Perceptually aware image inpainting

Abstract: Image inpainting is a process of reconstructing missing regions, or removing unwanted objects automatically by propagating intensity and texture information from surrounding parts of the image in a visually plausible manner. We propose a new exemplar-based image inpainting algorithm, which uses the recently developed metric called the perceptual-fidelity aware mean squared error (PAMSE). The PAMSE is a Gaussian-smoothed mean squared error (MSE) and approximates a weighted sum of the gradient of MSE, the Laplacian of MSE, and MSE itself. We show that, compared to MSE, PAMSE is a promising perceptual fidelity metric for application to image inpainting and leads to better performance in propagating texture and geometric structure simultaneously.

Size-Invariant Detection of Cell Nuclei in Microscopy Images

Abstract: Accurate detection of individual cell nuclei in microscopy images is an essential and fundamental task for many biological studies. In particular, multivariate fluorescence microscopy is used to observe different aspects of cells in cultures. Manual detection of individual cell nuclei by visual inspection is time consuming, and prone to induce subjective bias. This makes automatic detection of cell nuclei essential for large-scale, objective studies of cell cultures. Blur, clutter, bleed-through, imaging noise and touching and partially overlapping nuclei with varying sizes and shapes make automated detection of individual cell nuclei a challenging task using image analysis. In this paper we propose a new automated method for fast and robust detection of individual cell nuclei based on their radial symmetric nature in fluorescence in-situ hybridization (FISH) images obtained via confocal microscopy. The main contributions are two-fold. 1) This work presents a more accurate cell nucleus detection system using the fast radial symmetry transform (FRST). 2) The proposed cell nucleus detection system is robust against most occlusions and variations in size and moderate shape deformations. We evaluate the performance of the proposed algorithm using precision/recall rates, $F_{\beta }$ -score and root-mean-squared distance (RMSD) and show that our algorithm provides improved detection accuracy compared to existing algorithms.

Segmentation and detection of fluorescent 3D spots.

Abstract: The 3D spatial organization of genes and other genetic elements within the nucleus is important for regulating gene expression. Understanding how this spatial organization is established and maintained throughout the life of a cell is key to elucidating the many layers of gene regulation. Quantitative methods for studying nuclear organization will lead to insights into the molecular mechanisms that maintain gene organization as well as serve as diagnostic tools for pathologies caused by loss of nuclear structure. However, biologists currently lack automated and high throughput methods for quantitative and qualitative global analysis of 3D gene organization. In this study, we use confocal microscopy and fluorescence in-situ hybridization (FISH) as a cytogenetic technique to detect and localize the presence of specific DNA sequences in 3D. FISH uses probes that bind to specific targeted locations on the chromosomes, appearing as fluorescent spots in 3D images obtained using fluorescence microscopy. In this article, we propose an automated algorithm for segmentation and detection of 3D FISH spots. The algorithm is divided into two stages: spot segmentation and spot detection. Spot segmentation consists of 3D anisotropic smoothing to reduce the effect of noise, top-hat filtering, and intensity thresholding, followed by 3D region-growing. Spot detection uses a Bayesian classifier with spot features such as volume, average intensity, texture, and contrast to detect and classify the segmented spots aseither true or false spots. Quantitative assessment of the proposed algorithm demonstrates improved segmentation and detection accuracy compared to other techniques. © 2012 International Society for Advancement of Cytometry

Computer and Robot Vision

Abstract: Computer and Robot Vision Vol. 1, by R.M. Haralick and Linda G. Shapiro, Addison-Wesley, 1992, ISBN 0-201-10887-1.

Long-Range Directional Movement of an Interphase Chromosome Site Dependent on Actin and Nuclear Myosin

Abstract: Increasing evidence suggests functional compartmentalization of interphase nuclei. This includes preferential interior localization of gene-rich and early replicating chromosome regions versus peripheral localization of gene-poor and late replicating chromosome regions , association of some active genes with nuclear speckles or transcription "factories", and association of transcriptionally repressed genes with heterochromatic regions. Dynamic changes in chromosome compartmentalization imply mechanisms for long-range interphase chromatin movements. However, live cell imaging in mammalian cells has revealed limited chromatin mobility, described as "constrained diffusion". None of these studies, though, have examined a chromosome locus undergoing an inducible repositioning between two different nuclear compartments. Here we demonstrate migration of an interphase chromosome site from the nuclear periphery to the interior 1-2 hr after targeting a transcriptional activator to this site. Spot redistribution is perturbed by specific actin or nuclear myosin I mutants. Extended periods of chromosome immobility are interspersed with several minute periods in which chromosomes move unidirectionally along curvilinear paths oriented roughly perpendicular to the nuclear envelope at velocities of 0.1-0.9 microm/min over distances of 1-5 microm. Our results suggest an active mechanism for fast and directed long-range interphase chromosome movements dependent directly or indirectly on actin/myosin.

Robust Nucleus/Cell Detection and Segmentation in Digital Pathology and Microscopy Images: A Comprehensive Review

Abstract: Digital pathology and microscopy image analysis is widely used for comprehensive studies of cell morphology or tissue structure. Manual assessment is labor intensive and prone to interobserver variations. Computer-aided methods, which can significantly improve the objectivity and reproducibility, have attracted a great deal of interest in recent literature. Among the pipeline of building a computer-aided diagnosis system, nucleus or cell detection and segmentation play a very important role to describe the molecular morphological information. In the past few decades, many efforts have been devoted to automated nucleus/cell detection and segmentation. In this review, we provide a comprehensive summary of the recent state-of-the-art nucleus/cell segmentation approaches on different types of microscopy images including bright-field, phase-contrast, differential interference contrast, fluorescence, and electron microscopies. In addition, we discuss the challenges for the current methods and the potential future work of nucleus/cell detection and segmentation.

Superpixels: An evaluation of the state-of-the-art

Abstract: Superpixels group perceptually similar pixels to create visually meaningful entities while heavily reducing the number of primitives for subsequent processing steps. As of these properties, superpixel algorithms have received much attention since their naming in 2003 (Ren and Malik, 2003). By today, publicly available superpixel algorithms have turned into standard tools in low-level vision. As such, and due to their quick adoption in a wide range of applications, appropriate benchmarks are crucial for algorithm selection and comparison. Until now, the rapidly growing number of algorithms as well as varying experimental setups hindered the development of a unifying benchmark. We present a comprehensive evaluation of 28 state-of-the-art superpixel algorithms utilizing a benchmark focussing on fair comparison and designed to provide new insights relevant for applications. To this end, we explicitly discuss parameter optimization and the importance of strictly enforcing connectivity. Furthermore, by extending well-known metrics, we are able to summarize algorithm performance independent of the number of generated superpixels, thereby overcoming a major limitation of available benchmarks. Furthermore, we discuss runtime, robustness against noise, blur and affine transformations, implementation details as well as aspects of visual quality. Finally, we present an overall ranking of superpixel algorithms which redefines the state-of-the-art and enables researchers to easily select appropriate algorithms and the corresponding implementations which themselves are made publicly available as part of our benchmark at http://www.davidstutz.de/projects/superpixel-benchmark/ .

Nuclei Segmentation with Recurrent Residual Convolutional Neural Networks based U-Net (R2U-Net)

Abstract: Bio-medical image segmentation is one of the promising sectors where nuclei segmentation from high-resolution histopathological images enables extraction of very high-quality features for nuclear morphometrics and other analysis metrics in the field of digital pathology. The traditional methods including Otsu thresholding and watershed methods do not work properly in different challenging cases. However, Deep Learning (DL) based approaches are showing tremendous success in different modalities of bio-medical imaging including computation pathology. Recently, the Recurrent Residual U-Net (R2U-Net) has been proposed, which has shown state-of-the-art (SOTA) performance in different modalities (retinal blood vessel, skin cancer, and lung segmentation) in medical image segmentation. However, in this implementation, the R2U-Net is applied to nuclei segmentation for the first time on a publicly available dataset that was collected from the Data Science Bowl Grand Challenge in 2018. The R2U-Net shows around 92.15% segmentation accuracy in terms of the Dice Coefficient (DC) during the testing phase. In addition, the qualitative results show accurate segmentation, which clearly demonstrates the robustness of the R2U-Net model for the nuclei segmentation task.