Department of Computer Science & Engineering

&ldquo;IIT Jodhpur is committed to impart quality education to facilitate the creation of qualified, competent and responsible human resources to meet the emerging technological challenges of the world. It is working towards creating an equitable learning platform that is positive, accessible and effective in the face of the modern societal and scientific advancements.

Flexibility in curriculum, in all its academic programs, encompassing elements of teaching, research, creativity, innovation and entrepreneurship is the most unique feature of IIT Jodhpur. The Department of Science and Technology (DST), under the National Mission on Cyber Physical Systems, has identified IIT Jodhpur as a Technology &amp; Innovation Hub in the area of Augmented Reality and Virtual Reality. The institute has also been nominated as the coordinator for the Jodhpur City and Knowledge Cluster by the Office of Principal Scientific Adviser, Government of India.

The sprawling 852-acre IIT Jodhpur campus is a majestic tribute to the deep-rooted diverse cultural heritage of the region. The campus is an architectural masterpiece, which is slated to become an international exemplar of sustainability with net-zero energy, and advanced water and waste management strategies.&rdquo;

IIT Jodhpur

Large scale machine learning is becoming an active research area recently. Most of the existing clustering algorithms cannot handle big data due to its high time and space complexity. Among the clustering algorithms, eigen vector based clustering, such as Spectral clustering, shows very good accuracy, but it has cubic time complexity. There are various methods proposed to reduce the time and space complexity for eigen decomposition such as Nyström method, Lanczos method etc. Nyström method has linear time complexity in terms of number of data points, but has cubic time complexity in terms of number of sampling points. To reduce this, various Rank k approximation methods also proposed, but which are less efficient compare to the normalized spectral clustering. In this paper we propose a two step algorithm for spectral clustering to reduce the time complexity to O(nmk + m2k′), by combining both Nyström and Lanczos method, where k is the number of clusters and k′ is the rank k approximation of the sampling matrix (k &lt; k′ &lt; &lt; m &lt; &lt; n). It shows very good results, with various data sets, image segmentation problems and churn prediction of a telecommunication data set, even with very low sampling (for 10 Million x 10 Million matrix, sampled only 100 columns) with lesser time, which confirms the validity of the algorithm. © 2016 ACM.

World University root node

ACM International Conference Proceeding Series

Scalable clustering and applications

Automatic technique of 2D to 3D image conversion is proposed using manifold learning and sequential labeling which generates very reliable and accurate 3D depth maps that are very close to ground truth depths. In paper, LLE which is a non linear and neighborhood preserving embedding algorithm is used for depth estimation of a 2D image. And then, fixed point supervised learning algorithm is applied to construct consistent and smooth 3D output. The high dimensional data points or pixels of the input frames can be represented by a linear combination of its nearest neighbors and a lower dimensional point is reconstructed while preserving the local and geometric properties of the frames. The neighbors are assigned to each input point in the image data set and their weight vectors are computed that best linearly reconstruct the input point from its neighbors. To get the depth value of input point in new image, the reconstruction weights of its closest neighbors in training samples are multiplied with their corresponding ground truth depth values. The fixed point learning algorithm takes depths from manifold and other image features as input vectors and generates more consistent and accurate depth images for better 3D conversion. © 2015 IEEE.

2015 5th National Conference on Computer Vision, Pattern Recognition, Image Processing and Graphics, NCVPRIPG 2015

Automated conversion of 2D to 3D image using manifold learning

We discuss the signal tracking approach based phase estimation in digital holographic interferometry and its potential application for continuous evaluation of out-offplane thermal expansion of a metallic surface as a case study. © OSA 2016.

Optics InfoBase Conference Papers

Signal tracking approach based phase estimation for analysis of thermal expansion by digital holographic iwnterferometry

The need for image fusion in current image processing systems is increasing mainly due to the increased number and variety of image acquisition techniques. Image fusion is the process of combining substantial information from several sensors using mathematical techniques in order to create a single composite image that will be more comprehensive and thus more useful for a human operator or other computer vision tasks. This paper presents a new approach to multifocus image fusion based on sparse signal representation. Block-based compressive sensing integrated with a projection-driven compressive sensing (CS) recovery that encourages sparsity in the wavelet domain is used as a method to get the focused image from a set of out-of-focus images. Compression is achieved during the image acquisition process using a block compressive sensing method. An adaptive thresholding technique within the smoothed projected Landweber recovery process reconstructs high-resolution focused images from low-dimensional CS measurements of out-of-focus images. Discrete wavelet transform and dual-tree complex wavelet transform are used as the sparsifying basis for the proposed fusion. The main finding lies in the fact that sparsification enables a better selection of the fusion coefficients and hence better fusion. A Laplacian mixture model fit is done in the wavelet domain and estimation of the probability density function (pdf) parameters by expectation maximization leads us to the proper selection of the coefficients of the fused image. Using the proposed method compared with the fusion scheme without employing the projected Landweber (PL) scheme and the other existing CS-based fusion approaches, it is observed that with fewer samples itself, the proposed method outperforms other approaches. © 2016 Optical Society of America.

Journal of the Optical Society of America A: Optics and Image Science, and Vision

Adaptive multifocus image fusion using block compressed sensing with smoothed projected Landweber integration in the wavelet domain

Large scale biometrics projects rely on capturing images/signal from multiple sensors. For example, in India's Aadhaar project, multiple fingerprint sensors of different make and model are used for data collection. Similarly, in law enforcement applications, different agencies use different fingerprint sensors. These scenarios cause two potential problems: (i) sensor inter-operability and (ii) protecting/recording chain of evidence. While sensor inter-operability in fingerprints is a well studied problem, automatically recording chain of evidence is a relatively less explored research problem. For both the problems, one potential approach includes automatically identifying sensors based on the input image. This paper presents a novel fingerprint sensor identification algorithm based on multiple features such as Haralick, entropy, statistical and image quality features. The proposed algorithm is evaluated on a large database with 30,000 images with 15 fingerprint sensor classes. The proposed algorithm achieves an accuracy of 96% and computationally requires less than 10 milliseconds for an image. © 2016 IEEE.

Proceedings - International Conference on Pattern Recognition

Fingerprint sensor classification via Mélange of handcrafted features

CAPTCHA (Completely Automated Public Turing Test to tell Computers and Humans Apart) have been a common tool for preventing unauthorized access to websites for over a decade, but increasingly sophisticated optical character recognition algorithms and attack strategies have rendered traditional CAPTCHAs insecure. In this paper, we propose a new CAPTCHA incorporating multiple biometric modalities. Users are asked to identify faces, eyes, and fingerprints in a complex composite image. With over 1,900 volunteers and 30,000+ attempts, the proposed approach achieves high human accuracy while being resistant to existing attacks on CAPTCHAs and to detection by state-of-the-art software. © 2016 IEEE.

2016 IEEE 8th International Conference on Biometrics Theory, Applications and Systems, BTAS 2016

A multibiometrics-based CAPTCHA for improved online security

Increase in the popularity and reach of short text messages has led to their usage in propagating unsolicited advertising, promotional offers, and other unwarranted material to users. This has led to a high influx of such spam messages. In order to protect the interests of the user, several countermeasures have been deployed by telecommunication companies to hinder the volume of such spam.However, some volume of spam messages still manage to avoid these measures and cause varying degree of annoyance to users. In this chapter, an automated spam detection algorithm is proposed to deal with the particular problem of short text message spam. The proposed algorithm performs the two class (spam, ham) classification using stylistic and text features specific to short text messages. The algorithm is evaluated on three databases belonging to diverse demographic settings. Experimental results indicate that the proposed algorithm is highly accurate in detecting spam in short messages and can be utilized by awide variety of users to reduce the volume of spam messages. © Springer India 2016.

Advances in Intelligent Systems and Computing

Automated spam detection in short text messages

Classifier fusion is a well-studied problem in which decisions from multiple classifiers are combined at the score, rank, or decision level to obtain better results than a single classifier. Subsequently, various techniques for combining classifiers at each of these levels have been proposed in the literature. Many popular methods entail scaling and normalizing the scores obtained by each classifier to a common numerical range before combining the normalized scores using the sum rule or another classifier. In this research, we explore an alternative method to combine classifiers at the score level. The Pool Adjacent Violators (PAV) algorithm has traditionally been utilized to convert classifier match scores to confidence values that model posterior probabilities for test data. The PAV algorithm and other score normalization techniques have studied the same problem without being aware of each other. In this first ever study to combine the two, we propose the PAV algorithm for classifier fusion on publicly available NIST multi-modal biometrics score dataset. We observe that it provides several advantages over existing techniques and find that the interpretation learned by the PAV algorithm is more robust than the scaling learned by other popular normalization algorithms such as min-max. Moreover, the PAV algorithm enables the combined score to be interpreted as confidence and is able to further improve the results obtained by other approaches. We also observe that utilizing traditional normalization techniques first for individual classifiers and then normalizing the fused score using PAV offers a performance boost compared to only using the PAV algorithm. © 2016 IEEE.

Improving classifier fusion via Pool Adjacent Violators normalization

Grading of cancer offers insight to the occurrence and progress of the disease. The course of treatment is planned depending on the grade of cancer. Segmentation of the glandular structure of tissue is a prerequisite for grading of colon, prostate and breast cancers. Manual segmentation method is time-consuming and suffers from the curse of observer bias. We propose an automated solution for gland segmentation from hematoxylin & eosin (H&E) stained histology images. Our method relies on the biological cue rather than gland specific signatures that may vary across the slides. We construct a novel informative morphological scale space for gland segmentation. The scale space uses the entropy of the connected components in a novel manner to prevent over segmentation of objects. Our solution is fast, accurate and applicable in a clinical setup. Experiments show an average F1 score of 0.68 for 85 histology images in 20x magnification. We obtain ∼ 30% improvement in F1 score compared to the area morphological scale space method. © 2016 IEEE.

Proceedings - International Conference on Image Processing, ICIP

Gland segmentation from histology images using informative morphological scale space

Prolonged usage of illicit drugs alter texture and geometric variations of a face and hence, affect the performance of face recognition algorithms. This research proposes a two fold contribution for advancing the state-of-art in recognizing face images with variations caused due to substance abuse: firstly, scattering transform (ScatNet) based face recognition algorithm is proposed. The algorithm yields good results however, it is very expensive in terms of the computational time and space. Therefore, as the next contribution, an autoencoder-style mapping function (AutoScat) is proposed that learns to encode the ScatNet representation of a face image to reduce the computation time. The results are evaluated on the publicly available Illicit Drug Abuse Face database. The results show that ScatNet based face recognition algorithm outperforms two commercial matchers. Further, compared with ScatNet, AutoScat is able to achieve lower rank-1 accuracy but requires 10-3 times lesser computational requirements and around 400 times smaller feature space. © 2016 IEEE.