Envision: Intelligent, Reliable, Quality Solution

Slides:



Advertisements
Similar presentations
© 2008 Oracle Corporation – Proprietary and Confidential.
Advertisements

Patient information extraction in digitized X-ray imagery Hsien-Huang P. Wu Department of Electrical Engineering, National Yunlin University of Science.
Distributed Systems Architectures
Chapter 7 System Models.
STATISTICS Univariate Distributions
Spectral Analysis of Function Composition and Its Implications for Sampling in Direct Volume Visualization Steven Bergner GrUVi-Lab/SFU Torsten Möller.
International Telecommunication Union Workshop on Standardization in E-health Geneva, May 2003 Digital Imaging in Pathology for Standardization Yukako.
Environmental Remote Sensing GEOG 2021 Lecture 2 Image display and enhancement.
Environmental Remote Sensing GEOG 2021
Polygon Scan Conversion – 11b
Construction Automation A Six Year Plan in Development Ron Singh Chief of Surveys Geometronics Manager Design to Dozer August, 2010.
Dr. Marc Valliant, VP & CTO
Knowledge Extraction from Technical Documents Knowledge Extraction from Technical Documents *With first class-support for Feature Modeling Rehan Rauf,
Computer Literacy BASICS
Our Digital World Second Edition
Configuration management
Chapter 5 – Enterprise Analysis
1 Photometric Stereo Reconstruction Dr. Maria E. Angelopoulou.
Chapter 11: Models of Computation
CS525: Special Topics in DBs Large-Scale Data Management
1. Documents types Visas (ID-2) ICAO standard passports (ID-3) ID cards and driving licences (ID-1) Travel and identity documents.
McGraw-Hill/Irwin McGraw-Hill/Irwin Copyright © 2009 by The McGraw-Hill Companies, Inc. All rights reserved.
Pennsylvania Value-Added Assessment System (PVAAS) High Growth, High Achieving Schools: Is It Possible? Fall, 2011 PVAAS Webinar.
Reconstruction from Voxels (GATE-540)
Computer vision: models, learning and inference
Computer Graphics An Introduction. What’s this course all about? 05/10/2014 Lecture 1 2 We will cover… Graphics programming and algorithms Graphics data.
1 Motion and Manipulation Configuration Space. Outline Motion Planning Configuration Space and Free Space Free Space Structure and Complexity.
Technische Universität München Fakultät für Informatik Computer Graphics SS 2014 Sampling Rüdiger Westermann Lehrstuhl für Computer Graphik und Visualisierung.
DTU Informatics Introduction to Medical Image Analysis Rasmus R. Paulsen DTU Informatics TexPoint fonts.
By: Mani Baghaei Fard.  During recent years number of moving vehicles in roads and highways has been considerably increased.
Chapter 13: Digital Control Systems 1 ©2000, John Wiley & Sons, Inc. Nise/Control Systems Engineering, 3/e Chapter 13 Digital Control Systems.
People Counting and Human Detection in a Challenging Situation Ya-Li Hou and Grantham K. H. Pang IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART.
Adaptive Segmentation Based on a Learned Quality Metric
Road-Sign Detection and Recognition Based on Support Vector Machines Saturnino, Sergio et al. Yunjia Man ECG 782 Dr. Brendan.
Digital Image Forensics
Face Recognition & Biometric Systems, 2005/2006 Face recognition process.
Detecting Digital Image Forgeries Using Sensor Pattern Noise presented by: Lior Paz Jan Lukas, jessica Fridrich and Miroslav Goljan.
Overview of Computer Vision CS491E/791E. What is Computer Vision? Deals with the development of the theoretical and algorithmic basis by which useful.
Detecting Image Region Duplication Using SIFT Features March 16, ICASSP 2010 Dallas, TX Xunyu Pan and Siwei Lyu Computer Science Department University.
 Image Search Engine Results now  Focus on GIS image registration  The Technique and its advantages  Internal working  Sample Results  Applicable.
2007Theo Schouten1 Introduction. 2007Theo Schouten2 Human Eye Cones, Rods Reaction time: 0.1 sec (enough for transferring 100 nerve.
Evaluating the Quality of Image Synthesis and Analysis Techniques Matthew O. Ward Computer Science Department Worcester Polytechnic Institute.
Introduction What is “image processing and computer vision”? Image Representation.
Paul Blythe and Jessica Fridrich Secure Digital Camera.
Digital Image Processing In The Name Of God Digital Image Processing Lecture1: Introduction M. Ghelich Oghli By: M. Ghelich Oghli
Introduction to Multimedia Security Topics Covered in this Course Multimedia Security.
Multimedia Databases (MMDB)
Perception Introduction Pattern Recognition Image Formation
 In electrical engineering and computer science image processing is any form of signal processing for which the input is an image, such as a photograph.
Compiled By: Raj G Tiwari.  A pattern is an object, process or event that can be given a name.  A pattern class (or category) is a set of patterns sharing.
DIGITAL IMAGE PROCESSING
1 Digital Image Processing Dr. Saad M. Saad Darwish Associate Prof. of computer science.
Biometrics Stephen Schmidt Brian Miller Devin Reid.
MACHINE VISION Machine Vision System Components ENT 273 Ms. HEMA C.R. Lecture 1.
Exposing Digital Forgeries in Color Filter Array Interpolated Images By Alin C. Popescu and Hany Farid Presenting - Anat Kaspi.
1 Iris Recognition Ying Sun AICIP Group Meeting November 3, 2006.
Autonomous Robots Vision © Manfred Huber 2014.
Jack Pinches INFO410 & INFO350 S INFORMATION SCIENCE Computer Vision I.
1 Machine Vision. 2 VISION the most powerful sense.
IMAGE FORGERY DETECTION Submitted by Deepika Dileep Deepika Dileep S7 IT N0:35 N0:35.
By Kyle Bickel. Road Map Biometric Authentication Biometric Factors User Authentication Factors Biometric Techniques Conclusion.
1 INTRODUCTION TO COMPUTER GRAPHICS. Computer Graphics The computer is an information processing machine. It is a tool for storing, manipulating and correlating.
Optical Character Recognition
IMAGE PROCESSING is the use of computer algorithms to perform image process on digital images   It is used for filtering the image and editing the digital.
Visual Information Processing. Human Perception V.S. Machine Perception  Human perception: pictorial information improvement for human interpretation.
Digital Image Processing Introduction
Outline Announcement Texture modeling - continued Some remarks
Introduction to Multimedia Security Topics Covered in this Course
Presentation transcript:

Envision: Intelligent, Reliable, Quality Solution Mausumi Acharyya, Ph.D. CEO Advenio TecnoSys (P) Ltd

Executive Summary Key areas we specialize: We are a global firm delivering the best possible Image and Data Analysis services in different vertical markets. Advenio TecnoSys is involved in Data and Image Analytics. We are a multidisciplinary engineering system and software development team aimed at providing quality service and solutions to Image Processing needs. We have a highly skilled development team, well positioned with over 40 years of collective R&D experience. Image Analytics Image Forensic HPC Solutions Video Analytics Data Analytics

Image Processing & Vision Our Offering Image Analytics Video Analytics Big Data Analytics Forensic Analysis Image Processing & Vision Security Machine Learning Image Forensic ML for Big Data: Assessment of statistical methodology needs/ possibilities Custom ML modeling and algorithm design Scalability planning Implementation architecture and integration Security Proactive Solutions: Number plate recognition Occupancy detection Intelligent scene verification People counting Perimeter protection Camera tampering alarm Traffic flow monitoring Panic alarm Left object detection Image Forensic Applications: Image & Video Analysis - Examination, evaluation, analysis, enhancement, authentication, comparison Image & Video Authentication – validate authenticity Comparative Analysis - comparisons of people, clothing, or vehicles involved in crimes or accidents, or other objects of evidence Image Analysis Applications : Healthcare – Medical Images (CT, MRI, PET, X-ray etc.), microscopic images (histology, cytology) , molecular images (gene expression, microarray) data Life Science – Bioinformatics, Biotechnology Security – Biometrics (face, finger print, iris recognition) Industrial Imaging – Defect Identification Defense – Radar Images HPC Parallel Processing High Performance Computing Services: Design, Develop & Maintain Applications for large data set computation in real time using immense parallel processing capabilities of CPU/GPU, GPGPU

Research & Development We identify the specific cutting-edge technologies and need in healthcare domain and design and develop innovative solutions based on these needs We offer complete productization of research output at academic level and go for licensing We offer upgradation of existing products in absence of right resources and expertise at client’s end Productization Advenio Original algorithm development as per market need Modification of existing or published solutions Research & Development Technical Consultation Convert research prototype to deliverable product Innovative Solution Upgradation Consultation Product development consultation We provide a wide range of solutions and services that help organizations to transform their business

Hyperspectral & Multispectral Remote Sensing Mathematical Modeling Knowledge Base Image Processing Pattern Recognition Machine Learning Computer Vision Data Mining Hyperspectral & Multispectral Remote Sensing Statistical analysis Video Processing Mathematical Modeling Soft-Computing

Specialized Knowledge Object Recognition & Detection Shape Analysis Segmentation Texture Analysis & Synthesis Image Reconstruction Image & Video matching Image Compression & Decompression Feature extraction in spatial and frequency domain Data Fusion/Registration Pre- and Post- Processing (image corrections, noise removal, deblurring, image enhancement etc.) Classification and Prediction Video Processing

Domain Healthcare – Medical Images (CT, MRI, PET, X-ray etc.), microscopic images (histology, cytology) , molecular images (gene expression, micro-array) data Life Science – Bioinformatics, Biotechnology Security – Biometrics (face, finger print, iris recognition) Industrial Imaging – Defect/Purity Identification (Textile, Food, Consumable Items, Civil Structure, Mechanical Systems, Steel Plant, Coal Mines and many more) Video Processing - Number plate recognition, Occupancy detection, Intelligent scene verification, People counting, Left object detection, Camera tampering alarm, Traffic flow monitoring etc. Defense – Radar Images Agriculture – Satellite Images Mining - Satellite Images Climatology - Satellite Images Social & Media – Any kind of Image Processing for Fun Applications And Many More

Case Studies Credible Projects

Texture Analysis Considerable knowledge and experience on Texture Analysis Developed several novel algorithms for Texture Analysis Segmentation of multi-textured images Textures which are even difficult to discriminate visually by human eye were being discriminated by the machine vision algorithm developed Features developed were affine transformation (scale, rotation, translation) invariant Document image (scanned documents) segmentation for separating textual and graphics parts applicable for second generation compression Features developed were invariant to skewness, resolution, rotation etc. Satellite image segmentation for classifying different regions like: habitation, vegetation, water bodies, man-made structures, open areas etc. Designed and developed a novel neuro-fuzzy based unsupervised classifier for this purpose

Texture Analysis Algorithm involves in-depth knowledge of Texture Analysis, Wavelet Analysis, Feature Extraction, Classification, Fuzzy Set Theory, Neural Network & Soft Computing Input Textured image where some of the regions are not discriminable as different class even visually Segmented image based on different textured regions Input Textured image where some of the regions are not discriminable as different class even visually Segmented image based on different textured regions Novelty of the algorithm is that it identifies textured regions as different class which are even very difficult to identify visually by human

Document Segmentation – Second Generation Compression Skewed image and its segmented output Test image with document skewed and text regions with different orientations and its segmented output Test image with non-convex and overlapping object boundaries and its segmented output Novel algorithm developed for segmentation of textual part from graphics part which is essential for second generation compression. Here text and graphics parts were considered as two different textured regions.

Satellite Image Classification Several regions were accurately segmented by the algorithm based on textures. In the Bombay IRS image even the two different densities in the Arabian Sea were identified correctly. In the Calcutta IRS image the open space which is a race course and even the Howrah bridge were identified correctly.

Nodule (Lesion) Detection in Chest X-ray 4 US Patents Novel algorithms were developed for identification of nodules (lesions) in chest X-rays. Chest X-ray data is a 2D projection image where each pixel represents a volumetric integration posing a challenge in detection and estimation of nodules and their characteristics. Several algorithms were developed to reduce False Positives or Nodule-look-alikes detection. Algorithm involves in-depth knowledge of Segmentation, Mathematical modeling, Pre- & Post Processing, Fractal Analysis, Shape Analysis, Image Analysis, Machine Learning, Statistical Analysis

Bleed (Aneurysm) Detection in Brain CT, Trauma Cases 2 US Patents Novel algorithms were developed for identification of subtle bleeds in brain CT images without contrast for trauma cases posing immense challenge Several algorithms were developed to reduce False Positives bleed-look-alikes (calcification, fresh-flowing blood, bright falx) Algorithm involves in-depth knowledge of Segmentation, Registration, Mathematical modeling, Pre- & Post Processing, Morphological Operations, Image Analysis, Machine Learning

Lie Detection in Functional MRI (fMRI) Images Novel approach one of the first attempts to find alternative to Polygraphs. Novel algorithm for identifying whether someone is telling a lie or truth based on functional MRI images. A one hundred percent classification accuracy was obtained on a set of unknown data (images) Algorithm involves exhaustive use of Registration both Rigid & Deformable and Non-linear Pattern Classification

Tuberculosis Bacilli Detection in Sputum Smears (microscopic image) US Patent Algorithm identifies TB Bacilli in microscopic images of sputum smears Algorithm based on Color Image analysis, Morphology, Machine Learning

Hairline Crack Detection in Bone X-ray, CT etc US Patent Algorithm involves use of exhaustive Image analysis like several feature extraction including fractal analysis, Gabor orientation mapping etc. & Machine Learning Techniques

Foreign Material and Crack Detection in Chocolate Bars (X-ray) Original image Crack identification Original image Identified foreign element Defect Analysis in X-ray images of Chocolate bars This involved an unsupervised technique to identify crack or presence of any foreign materials inside the bars. In this analysis no reference image were used and it is completely a machine vision technique Original image Identified foreign element

Segmentation of stained microscopic image of Root Segmentation of stained microscopic images of Root Algorithm involves different state-of-art image segmentation techniques

Pulmonary Embolism Detection in CT Angiogram Pulmonary embolism (PE) is a serious medical condition, characterized by the partial/complete blockage of an artery within the lungs Novel algorithm developed for PE detection and several algorithms for reducing False Positives or PE-look-alikes Algorithm involves use of exhaustive Image analysis & Machine Learning Techniques

Subtle Nodule Detection in Low Resolution Lung CT images Algorithm identified very subtle nodules which were missed by radiologists Algorithm involves use of exhaustive Image analysis, Deformable Registration & Machine Learning Techniques

Alzheimer Detection in Diffusion Tensor MRI (DTMRI) Images Diffusion tensor tractography Algorithm based on Singular Value Decomposition, Tensor Analysis, Deformable Registration and intensive Mathematical modeling

Mammogram Segmentation and Nodule Detection Mammogram segmentation and Nodule detection. It is extremely challenging to segment mammograms which are digitized from X-ray films due to presence of noise and other artifacts Algorithm based on Morphological operations, Segmentation based on Active Contour and Fuzzy Sets and intensive Mathematical modeling

Bitting Code Generation in Locksmith Key Locksmith Key Coding: Mobile Application Automatic detection, segmentation, bit edge profiling and generation of code from space and depth information

Face Authentication Face Authentication: Face recognition feature extraction, authentication

Iris Authentication Iris Authentication: Eye lash and other denoising and segmentation of pupil and iris region. Encoding the iris for authentication

Optical Character Recognition (OCR) Automated Answer Script Evaluation (Optical Character Recognition) Algorithm involves rigid registration and Scale Invariant Feature Transform (SIFT)

Optical Character Recognition Removal of “Phrase” from Text Document

Bioacoustics Bird Chirp Identification and Classification

Accelerometer Data Analysis Golf Swing Key Points Identification

Watermark Watermarking - novel algorithm developed to embed patient information in medical images as watermark The uniqueness and novelty of the algorithm lies in the fact that apart from the use of embedded watermark as an information source we can watermark some extra ‘text’ information pertaining to the image. We see this as a revolutionary tool in the arsenal of a radiologist using a CAD suite, wherein he can embed certain crucial information into an image rather than relying on the DICOM header. The significant benefit would be the selective access of this information and its being free of traditional viewing platforms. Thus the watermark here acts in actual way of both conveying information as well as providing security. A failure of the watermark would mean the non-reliability of both the image as well as the additional information thereby maintaining its traditional purpose. Algorithm involves Wavelet Analysis, quantization, coding Published Patent

Satellite Image Segmentation Region classification of satellite images into different categories like vegetation, building area, shadows, road area etc

Satellite Image Segmentation

Satellite Image Segmentation

Liver Histology (Biopsy) Image Analysis for Automatic Detection and Grading of Liver Diseases Automatic Lesion detection and gradation of Liver diseases Algorithm involves novel machine learning and image processing techniques for extraction of features and classification Steatosis Lobular Inflammation

Histology (Microscopic Tissue Images) Image Analysis of different Organs for aiding Drug Development Detection of Normal and Abnormal Organs Algorithm involves novel machine learning and image processing techniques for extraction of features and classification

Image Forensic Identifying image tampering, alteration or forgery Passive Authentication - Patterns of original images are distinguishable due to the different imaging devices and image processing inside them. Original patterns constrained by the statistical characteristics in natural scenes, physical conditions in a scene, etc. would be altered after tampering. Passive forensics can be converted to a problem of pattern recognition Solution to the problem is finding the different patterns according to the knowledge from various imaging devices or the manipulations or the natural scene constraints, etc. Selected patterns with distinguishing ability are crucial for this new technology

Knowledge of image acquisition model Passive Forensic Identification of image source Imaging devices have varying characteristics due to different physics apparatus image processing parameters applied inside the imaging devices Lead to different patterns of the output images Use these patterns as inherent “fingerprints” of the imaging devices to identify the source of the image Feature Extraction Feature Map Pattern Extraction Pattern Map Image X Imaging Device Knowledge of image acquisition model Compare Similarities between patterns and characteristic features measured. Confidence measure for each imaging device to identify the source of the image is computed.

Knowledge of image manipulation model Passive Forensic Detection of image alteration Original images always contain some consistent characteristics noise distribution light condition Characteristics change after image post-processing (alterations) Some features of the altered images become more or less inconsistent Finding the differences before and after the operations is the key of the technology Feature Extraction Feature Map Pattern Extraction Pattern Map Image X Original/ Altered Image Knowledge of image manipulation model Compare Extract features from image X and obtain the original/ altered patterns mainly using the knowledge of the image manipulation model. Compare the distance between the features and the patterns to decide whether or not image X has been altered

Image Forensic Fourier analysis of histograms of the images to determine if a JPEG image underwent double compression Localize which portions underwent double compression by analyzing difference images Any forgery in the image or splicing of two images from different cameras can be identified by analyzing the correlations introduced by performing color filter array (CFA) interpolation and demosaicking Any tampering can be measured by the local correlation of camera noise or photo-response non-uniformity noise (PRNU) and the image Forgery due to splicing of images require resampling which disturbs the specific correlation of the image. Estimating this correlation using expectation maximization can identify the forgery A common manipulation is to copy and paste portions of an image to replicate an object or conceal a person in a scene called cloning. Presence of identical (or virtually identical) regions in an image can, therefore, be used as evidence of tampering. Image similarity measure, by matching key points in the images using different techniques like scale invariant feature transform (SIFT) and random sample consensus algorithm (RANSAC) can be used to identify this cloning

Image Forensic Computer graphics (CG) capable of generating highly realistic images. But CG images are rendered under idealized models of geometry, lighting, camera optics and sensors, and their underlying statistics differ from those of photographic images. Image decomposition at different orientation, scale and spatial frequency that are localized in spatial position would capture all the statistical irregularities in case of CG images Creating a photo composite requires translation, scaling, rotations of portions of an image. Such image-based manipulations can change the effective intrinsic camera parameters. Differences in these estimated parameters can be used as evidence of tampering Shadow analysis (light travels in straight line so lines drawn on shadows and objects should intersect at one point (light source). This can be used to analyze any forgery Image created by splicing together individual images is difficult to exactly match the lighting effects due to directional lighting (e.g., the sun on a clear day). Differences in lighting can be a telltale sign of tampering. Direction of the light source can be estimated for different objects/people in an image, inconsistencies in the lighting direction can be used as evidence of tampering

Products OptiNio – Retinal abnormality detection software CheckCount – Automated counter of production of any item

Brief Profile of Founder & CEO Dr. Mausumi Acharyya Post-Doc (Medical Imaging), University of Pennsylvania, USA PhD, Computer Science (Image Processing), Indian Statistical Institute, India M.Tech and B.Tech (Electronics & Telecommunication), Calcutta University, India 13+ years in IT industry (GE Global Research, Siemens, Defense Research & Development Organization) 20+ years of rich research experience in Multifarious domains Distinguished research and development exposure in Signal/Image processing Medical Image Processing Machine Learning and Vision In-depth understanding and extensive knowledge of designing and development of Image, Video, Data analytics applications Key in successfully delivering several commercial products during her engagement with the above organizations Mausumi holds several US Patents and international publications to her credit  

Advenio TecnoSys Pvt. Ltd. STEP, IIT, Kharagpur, India Contacts Advenio TecnoSys Pvt. Ltd. STEP, IIT, Kharagpur, India Dr. Mausumi Acharyya, CEO +918607026111, +91 9845311644 mausumi.acharyya@adveniotecnosys.com info@adveniotecnosys.com

Proprietary and Confidential Thank You Proprietary and Confidential By accepting this document the recipient agrees to keep all information permanently confidential or otherwise acquired by the receiving party from Advenio TecnoSys including all handbooks, manuals, drawings, designs, specifications, charts, diagrams etc. and any other documents or materials containing such information. The information presented in this document is for discussion purposes only and is subject to change. Advenio TecnoSys shall not be liable for errors contained herein, or for consequential damages in connection with the furnishing, performance or use of the material. No part of this document may be reproduced or transmitted in any form or by any means, for any purpose, or translated to another language without the prior written consent of Advenio TecnoSys.

Transpiring Innovation Thank You Transpiring Innovation