Research

3D cell segmentation and analysis of coordinated in cell division (Coming Soon !)

Cell migration : Shape-time series analysis (Coming Soon !)

Multi-domain Multi-task cell segmentation (Coming Soon !)

Saliency aided dictionary learning for image similarity

In image classification, obtaining adequate data to learn a robust classifier has often proven to be difficult in several scenarios. To adequately exploit the limited training data in classification, a saliency guided dictionary learning method and subsequently an image similarity technique was proposed for histo-pathological image classification. A dictionary learning algorithm was developed by leveraging the salient regions in an image. The sparse representation of an image with respect to the dictionary learned from another is analyzed to quantify the similarity between images.

Publications:

1. SDL: Saliency guided dictionary learning for image similarity, IEEE TIP 2018 (pdf, link, code)

2. SLIDE: Saliency guided image dictionary and image similarity evaluation, ICIP 2016. (pdf , link)

Rare event detection

Videos captured using handheld camera introduces jitter and notable changes in background. Thus background subtraction method fails to detect such occurrences. We propose to determine the significance of a frame, while preserving its compact representation, by introducing a saliency-driven dictionary learning technique.

Publications:

1. SSPARED: Saliency and sparse code analysis for rare event detection in video, IVMSP , 2016. (link)

2. GraDED: A graph-based parametric dictionary learning algorithm for event detection, ICIP 2017 (link)

Segmentation in presence of intensity inhomogeneity

Image segmentation in presence of Intensity inhomogeneity is a challenging task. This project involves segmentation of blood vessel boundary from ultrasound to assist medical practitioners for performing phlebotomy application such as intravenous needle placement. The proposed solution involves the integration of a level set segmentation methodology with the dictionary learning framework. This provides an elegant solution to deal with intensity inhomogeneities prevalent in many imaging applications such as ultrasound and fluorescence microscopy.

Publications:

1. Dictionary Learning Level Set, IEEE Signal Processing Letters, 2015. (pdf , link, code)

BACT 3D

Bact-3D, a 3D method for segmenting super-resolution images of multi-leveled, living bacteria cultured in vitro. The method incorporates a novel initialization approach that exploits the geometry of the bacterial cells as well an iterative local level set evolution that is tailored to the biological application.

Publication: Bact-3d: A level set segmentation approach for dense multi-layered 3d bacterial biofilms, ICIP 2017 (link)

Image classification by query adaptive feature selection

With increasing complexity of the dataset it becomes impractical to use a single feature to characterize all constituent images. This method automatically selects the image features, relevant for classification, without any modification to the feature extracting methods or the classification algorithm.

Publications:

1. 1. A meta-algorithm for classification by feature nomination, ICIP, 2014. (pdf , link)

2. 2. Image classification by multi-kernel dictionary learning, Asilomar, 2014. (pdf , link)

Tracking heliotropic motion of sunflower

In this work, an algorithm was developed to automatically track circum-nutation of sunflower seedlings. The tracking method uses active contour models with constraints on affine transformation for predicting the translated and rotated boundary of the sunflower leaves

Publications: Tracking Sunflower Circumnutation using Affine Parametric Active Contours, SSIAI, 2014. (pdf , IEEE Xplore)

Visual tracking in presence of illuminations variation

This work presents an algorithm for tracking moving objects across spatio-temporally varying illumination from a sequence of nonlinear observations corrupted by non-Gaussian noise. The key idea of the algorithm is to importance sample for the small dimensional state vector i.e, global motion, and constrained posterior mode tracking for recovering the sparse spatial signal i.e., illumination change.

Publications:

1. Tracking sparse signal sequences from nonlinear/non-Gaussian measurements and applications in illumination-motion tracking, ICASSP 2013. (pdf , IEEE Xplore)

2. PaFiMoCS: Particle Filtered Modified-CS and Applications in Visual Tracking across Illumination Change. (arXiv)