Steven Lawrence Fernandes

A comparative study on ICA and LPP based Face Recognition under varying illuminations and facial expressions

Dimensionality reduction has been a key problem in Face Recognition. Independent Component Analysis (ICA) is a recent approach for dimensionality reduction. Locality Preserving Projections (LPP) is also a recently proposed new method in pattern recognition for feature extraction and dimension reduction. In this paper we have developed and analyzed the face recognition rate of ICA and LPP under varying illuminations and facial expressions. Analyzes is performed on YALEB databases which contains 64 illuminations conditions (5760 images) and ATT databases which contains major facial expressions (400 images). From the results we conclude that the best algorithm to recognize images with varying illuminations is ICA. On the other hand to recognize image with varying facial expressions, LPP is better to use because it has better recognition rate.

Entropy based segmentation of tumor from brain MR images a study with teaching learning based optimization

This work proposes the meta-heuristic approach assisted segmentation and analysis of glioma from brain MRI dataset.A novel two stage approach is implemented based on tri-level thresholding and level set segmentation.A detailed analysis of well known entropy approaches, such as Kapur, Tsallis and Shannon are presented.A comparative study between level set and active contour segmentation is presented. Image processing plays an important role in various medical applications to support the computerized disease examination. Brain tumor, such as glioma is one of the life threatening cancers in humans and the premature diagnosis will improve the survival rate. Magnetic Resonance Image (MRI) is the widely considered imaging practice to record the glioma for the clinical study. Due to its complexity and varied modality, brain MRI needs the automated assessment technique. In this paper, a novel methodology based on meta-heuristic optimization approach is proposed to assist the brain MRI examination. This approach enhances and extracts the tumor core and edema sector from the brain MRI integrating the Teaching Learning Based Optimization (TLBO), entropy value, and level set / active contour based segmentation. The proposed method is tested on the images acquired using the Flair, T1C and T2 modalities. The experimental work is implemented and is evaluated using the CEREBRIX and BRAINIX dataset. Further, TLBO assisted approach is validated on the MICCAI brain tumor segmentation (BRATS) challenge 2012 dataset and achieved better values of Jaccard index, dice co-efficient, precision, sensitivity, specificity and accuracy. Hence the proposed segmentation approach is clinically significant. Display Omitted

Low Power Affordable and Efficient Face Detection in the Presence of Various Noises and Blurring Effects on a Single-Board Computer

Till today face detection is a burning topic for the researchers. In the areas like digital media, intelligent user interface, intelligent visual surveillance and interactive games. Various noises and blurring effects face images captured in real time. Single board computer for efficient face detection system is introduced in this paper which works well in the presence of Gaussian Noise, Salt & Pepper Noise, Motion Blur and Gaussian Blur. Raspberry Pi based single-board computer is used for the experiments, because it consumes less power and is available at an affordable price. The developed system is tested by introducing varying degree of noises and blurring effects on standard public face databases: GRIMACE, JAFEE, INDIAN FACE, CALTECH, FACE 95, FEI – 1, FEI – 2. In the absence of noise and blurring effects also the system is tested using standard public face databases: GRIMACE, JAFEE, INDIAN FACE, CALTECH, FACE 95, FEI – 1, FEI – 2, HEAD POSE IMAGE, SUBJECT, and FGNET. The key advantage of the proposed system is excellent face detection rates in the presence of noises, blurring effects and also in the presence of varying facial expressions and across age progressions. Python scripts are developed for the system, resulted are shared on request.

A method for the detection and classification of diabetic retinopathy using structural predictors of bright lesions

Abstract Diabetic burden around the world with a consequence of diabetic retinopathy can lead to permanent blindness in patients. Exudates detection in fundus images through an automated method is a vital task that has many applications in diabetic retinopathy screening. Realizing it important, a system being proposed in this paper automatically classifies exudates and non-exudates regions in retinal images. Presented technique is based on pre-processing for candidate lesion extraction, features extraction and classification. In pre-processing, Gabor filter is applied to the gray scale image which makes it useful for lesion enhancement. Segmentation of candidate lesion is based on mathematical morphology. A features set is selected for each candidate lesion using a combination of statistical and geometric features. Presented method is evaluated via publicly accessible datasets with the help of performance parameters such as true positive, false positive and area under curve for statistical analysis. Publicly available datasets such as e-ophtha, HRIS, MESSIDOR, DIARETDB1, VDIS, DRIVE, HRF and one local dataset are used to test the suggested system. The achieved results show an average AUC of 0.98 and accuracy as high as 98.58% which are substantially higher than the existing methods.

Recognizing Faces When Images Are Corrupted by Varying Degree of Noises and Blurring Effects

Most images are corrupted by various noises and blurring effects. Recognizing human faces in the presence of noises and blurring effects is a challenging task. Appearance based techniques are usually preferred to recognize faces under different degree of noises. Two state of the art techniques considered in our paper are Locality Preserving Projections (LPP) and Hybrid Spatial Feature Interdependence Matrix (HSFIM) based face descriptors. To investigate the performance of LPP and HSFIM we simulate the real world scenario by adding noises: Gaussian noise, Salt and pepper noise and also adding blurring effects: Motion blur and Gaussian blur on six standard public face databases: IITK, ATT, JAFEE, CALTECH, GRIMANCE, and SHEFFIELD.

Extraction of emotions from multilingual text using intelligent text processing and computational linguistics

Abstract Extraction of Emotions from Multilingual Text posted on social media by different categories of users is one of the crucial tasks in the field of opining mining and sentiment analysis. Every major event in the world has an online presence and social media. Users use social media platforms to express their sentiments and opinions towards it. In this paper, an advanced framework for detection of emotions of users in Multilanguage text data using emotion theories has been presented, which deals with linguistics and psychology. The emotion extraction system is developed based on multiple features groups for the better understanding of emotion lexicons. Empirical studies of three real-time events in domains like a Political election, healthcare, and sports are performed using proposed framework. The technique used for dynamic keywords collection is based on RSS (Rich Site Summary) feeds of headlines of news articles and trending hashtags from Twitter. An intelligent data collection model has been developed using dynamic keywords. Every word of emotion contained in a tweet is important in decision making and hence to retain the importance of multilingual emotional words, effective pre-processing technique has been used. Naive Bayes algorithm and Support Vector Machine (SVM) are used for fine-grained emotions classification of tweets. Experiments conducted on collected data sets, show that the proposed method performs better in comparison to corpus-driven approach which assign affective orientation or scores to words. The proposed emotion extraction framework performs better on the collected dataset by combining feature sets consisting of words from publicly available lexical resources. Furthermore, the presented work for extraction of emotion from tweets performs better in comparisons of other popular sentiment analysis techniques which are dependent of specific existing affect lexicons.

Recognizing facial images using ICA, LPP, MACE Gabor Filters, Score Level Fusion Techniques

We have developed and analyzed Independent Component Analysis (ICA), Locality Preserving Projections (LPP), Minimum Average Correlation Energy (MACE) Gabor Filters, Score Level Fusion Techniques (SLFT) for Face Recognition in the presence of various noises and blurring effects. ICA considers statistical characteristics in second order or higher order. LPP is used to generate an unsupervised neighborhood graph on training data, and then finds an optimal locality preserving projection matrix under certain criterion. MACE Gabor filter synthesizes a filter using a set of training images that would produce correlation output that minimizes correlation values at locations other than the origin and the value at the origin is constrained to a specific peak value. ICA, LPP, MACE Gabor Filter, SLFT are the 4 systems developed which were trained in the absence of noise, blurring effect but tested by imposing different levels of noises and blurring effects. To compare the performances six public face databases: IITK, ATT, JAFEE, CALTECH, GRIMACE, and SHEFFIELD are considered.

Robust Face Recognition in the Presence of Noises and Blurring Effects by Fusing Appearance Based Techniques and Sparse Representation

In real life, images obtained from video cameras or scanners are usually exposed to different levels of noises and blurring effects. In this paper we propose a new robust score level fusion technique to recognize faces in the presence of noise and blurring effects. The Proposed Score Level Fusion Technique (PSLFT) is obtained by using combinatory approach and Z-Score normalization using the scores obtained from appearance based techniques: Principal Component Analysis (PCA), Fisher faces (FF), Independent Component Analysis (ICA), Fourier Spectra (FS), Singular Value Decomposition (SVD) and Sparse Representation (SR). The system is trained in the absence of noise, blurring effect but tested by imposing different levels of noises and blurring effects thus we have tried to imitate the real world scenarios. To investigate the performance of PSLFT, we simulate the real world scenario by adding noises: Median noise, Salt and pepper noise and also adding blurring effects: Motion blur and Gaussian blur. To evaluate performance of the PSLFT, we have considered six standard public face databases: IITK, ATT, JAFEE, CALTECH, GRIMANCE, and SHEFFIELD.

ODROID XU4 based implementation of decision level fusion approach for matching computer generated sketches

Abstract Implementing computer vision applications on energy efficient and powerful single board computer devices is a hot topic of research. ODROID-XU4 is one such latest single board computing device which is extremely energy efficient and powerful, having a small form factor when compared to any other ARM based embedded devices. It supports open source operations systems and runs a variety of Linux flavors including Ubuntu and various Android versions including Lollipop. Moreover, it supports USB 3.0, eMMC 5.0 and Gigabit Ethernet interfaces thus, making the device feasible to transfer data at a very high speed. The key contribution of this paper is we have developed a novel technique to match computer generated sketches with face photos and implemented it on ODROID XU4 single board computer which makes it feasible to be used in real-time. Human face is detected on the face photos using Viola Jones method. On the detected faces and computer generated sketches, feature extraction is performed using supervised auto-encoder to build deep architecture and matching is performed between computer generated sketches and face photos using Parallel Convolutional Neural Network (PCNN). Finally decision level fusion is performed to find the optimal matching result. In this study, the authors have performed pilot testing of their technique and results of their analysis are presented to the readers.

A Comparative Study on Score Level Fusion Techniques and MACE Gabor Filters for Face Recognition in the Presence of Noises and Blurring Effects

Face recognition has been an intensely researched field of computer vision for the past couple of decades. Though significant strides have been made in tackling the problem in controlled domains, significant challenges remain in solving it in the unconstrained domain. Two such scenarios are while recognizing faces acquired from distant cameras and when images are corrupted. The main factors that make this a challenging problem are image degradation due to noise and blur. In this paper we have developed and analyzed Score Level Fusion Technique (SLFT) of appearance based techniques and Minimum Average Correlation Energy (MACE) Gabor filter for face recognition in the presence of various noises and blurring effects. In SLFT the scores are obtained by using combinatory approach and Z-Score normalization of appearance based techniques: Principal Component Analysis (PCA), Fisher faces (FF), Independent Component Analysis (ICA), Fourier Spectra (FS), Singular Value Decomposition (SVD) and Sparse Representation (SR). MACE Gabor filter is designed to minimize the average correlation energy (ACE) of the correlation outputs due to the training images while simultaneously satisfying the correlation peak constraints at the origin. The effect of minimizing the ACE is that the resulting correlation planes would yield values close to zero everywhere except at the location of a trained object, where it would produce a strong peak. We simulate the real world scenario by adding noises: Median noise, Salt and pepper noise and also adding blurring effects: Motion blur and Gaussian blur. To compare the performance of SLFT and MACE Gabor filter, we have considered six standard public face databases: IITK, ATT, JAFEE, CALTECH, GRIMANCE, and SHEFFIELD.