Prashant P. Patavardhan

Super-resolution for iris feature extraction

Super-resolution technique can be used to fix the low resolution problem for recognizing the iris at a distance. Two frequency domain super-resolution algorithms, Papoulis-Gerchberg (PG) and Projection onto Convex Sets, are implemented to increase the resolution of iris images. The performance analysis of these algorithms is carried out by extracting Gray Level Co-occurrence Matrix (GLCM) features of super-resoluted iris images. The super-resoluted iris region is normalized, extracted GLCM features and compared with the GLCM features of normalized original iris region. It has been observed that the GLCM features reconstructed images using above algorithm closely matches with that of original iris image. The error between the GLCM features of original normalized and normalized super-resoluted image using Papoulis-Gerchberg is less compared to that of Projection onto Convex Sets.

Super resolution and recognition of long range captured multi-frame iris images

In this study, a framework to super resolve and recognise the long range captured iris polar images is proposed. The proposed framework consists of best frame selection algorithm, modified diamond search algorithm, Gaussian process regression (GPR) based and enhanced iterated back projection (EIBP)-based super-resolution approach, fuzzy entropy-based feature selector and neural network (NN) classifier. The framework uses linear kernel co-variance function in local patch-based GPR and EIBP algorithms and it super resolves the iris images depending on the contents of the patches, without an external dataset. NN classifier classifies the iris images by using features extracted using discrete cosine transform domain based no-reference image quality assessment model, Gray level co-occurrence matrix, Hu seven moments and statistical features. The framework is tested using CASIA long range iris database by comparing and analysing the peak signal-to-noise ratio, structural similarity index matrix and visual information fidelity in pixel domain of proposed algorithms with Yang and Nguyen framework. The results demonstrate that the proposed framework is well suited for recognition of iris images captured at a long distance.

Super resolution based low cost vision system

Machine vision (MV) is the technology which provides camera based analysis of images for various applications such as automatic quality inspection, pattern recognition, process flow control and pattern classification. The machine vision system is expensive as it contains high resolution camera and lenses. The paper proposes an algorithm to develop a low cost web camera based vision system for screw thread inspection. The Bayesian super-resolution method is used to super-resolute the images captured using low resolution web cameras. The parameters such as major, minor and pitch diameters, depth and thread angles are measured by using the proposed dimension measurement method. The results of web camera based automatic inspection of major diameter, minor diameter, pitch diameter, thread and depth of hex lag screw thread shows an error of range 0.000 to 0.310 mm. The comprehensive experimental results reveal that the proposed approach is suitable for real-time high speed quality analysis in various industries.

Analysis of Face Recognition Algorithms for Uncontrolled Environments

Face recognition is a challenging problem in biometric systems, which has received a lot of attention in the last two decades as it has numerous applications in computer vision and pattern recognition. There is remarkable progress in the face recognition systems under controlled conditions, but they degrade for uncontrolled conditions like pose, illumination, expression, and occlusion etc. In this paper, we discussed different algorithms like PCA, DCT, LDA, ANN, ICA, HMM, and Wavelet with its pros and cons. The different face database used for face recognition is discussed. It also discusses various challenges and possible future directions for face recognition task.

Feature Extraction and Fuzzy-Based Feature Selection Method for Long Range Captured Iris Images

Long range captured iris recognition system is a biometric system consisting of pattern recognition and computer vision. In the process of iris recognition, feature extraction and feature selection play a major role in increasing the recognition accuracy. This paper proposes feature extraction method using discrete cosine transform domain-based no-reference image quality assessment model, gray-level co-occurrence matrix, Hu seven moments, and statistical features. It also proposes fuzzy entropy and interval-valued fuzzy set measure-based feature selection method. The selected feature vectors are classified by neural network classifier. The model is tested with CASIA long range iris database. The recognition accuracy is compared with the results obtained without feature selection and existing feature selection methods. It has been observed that the fuzzy entropy method gives better classification accuracy than existing feature selection method. The results demonstrate that the proposed work is well suited to extract the features of iris polar images captured at a long distance and to reduce the dimensionality by selecting the useful features which increase the recognition accuracy .

Unconstrained Iris Image Super Resolution in Transform Domain

In this paper, a method for super resolution of unconstrained or long-range captured iris images in discrete cosine transform domain is proposed. This method combines iterated back projection approach with the Papoulis-Gerchberg (PG) method to super resolute the iris images in discrete cosine transform domain. The method is tested on CASIA long-range iris database by comparing and analyzing the structural similarity index matrix, peak signal-to-noise ratio, visual information fidelity in pixel domain, and execution time of bicubic, Demirel, and Nazzal state-of-the-art algorithms. The result analysis shows that the proposed method is well suited for super resolution of unconstrained iris images in transform domain.

Gaussian Process Regression based iris polar image super resolution

In this work, Gaussian Process Regression (GPR) based novel framework is proposed to super resolute the long range captured iris polar images. The framework uses linear kernel co-variance function in GPR during the process of super resolution of iris image, without external dataset. The new technique is proposed to reduce the time taken to super resolute the iris polar image patches. The framework is tested using benchmark images as well as CASIA long range iris database by comparing and analyzing the peak signal to noise ratio (PSNR) and structural similarity index matrix (SSIM) of proposed algorithms with the existing algorithms. Empirical results indicate that the proposed framework, which improves PSNR up to 36 dB and promotes structural similarity index measurement (SSIM) up to 0.92 in averages, is better than the other existing method. The results demonstrate that the proposed approach outperforms some of the state-of-the-art super resolution approaches.

SINGLE FRAME SUPER RESOLUTION OF NONCOOPERATIVE IRIS IMAGES

Image super-resolution, a process to enhance image resolution, has important applications in biometrics, satellite imaging, high definition television, medical imaging, etc. The long range captured iris identification systems often suffer from low resolution and meager focus of the captured iris images. These degrade the iris recognition performance. This paper proposes enhanced iterated back projection (EIBP) method to super resolute the long range captured iris polar images. The performance of proposed method is tested and analyzed on CASIA long range iris database by comparing peak signal to noise ratio (PSNR) and structural similarity index (SSIM) with state-of-the-art super resolution (SR) algorithms. It is further analyzed by increasing the up-sampling factor. Performance analysis shows that the proposed method is superior to state-of-the-art algorithms, the peak signal-tonoise ratio improved about 0.1-1.5 dB. The results demonstrate that the proposed method is well suited to super resolve the iris polar images captured at a long distance.

Analysis of Skin Inheritance Using FOS

The texture and colour of human skin varies from one individual to another. These two features pave the way for study of skin in detail. These features are inherited by a child, either from its parents or grandparents or both. The study concentrates in determining the skin colour and texture dominance in a child by using image processing techniques. The colour spaces used for extraction of these features are RGB, HSV and YCbCr. First order statistical measures such as mean, standard deviation, variance and entropy are calculated of the captured images. The Fuzzy Logic Toolbox from MATLAB is used for classification and analysis of ambiguity in skin colour and textural features for genetic dominance classification.