Mudassar Raza

Illumination normalization preprocessing for face recognition

This paper describes an illumination normalization technique which works at the pre-processing stage where the face image is first divided into equal sub-regions. Each sub-region is then processed separately for illumination normalization. Then the segments are joined back followed by further processing like noise removal and contrast enhancement. The proposed technique is tested on Yale dataset and compared with some previous illumination normalization techniques.

Robust Face Recognition Technique under Varying Illumination

Face recognition is one of a complex biometrics in the field of pattern recognition due to the constraints imposed by variation in the appearance of facial images. These changes in appearance are affected by variation in illumination, expression or occlusions etc. Illumination can be considered a complex problem in both indoor and outdoor pattern matching. Literature studies have revealed that two problems of textural based illumination handling in face recognition seem to be very common. Firstly, textural values are changed during illumination normalization due to increase in the contrast that changes the original pixels of face. Secondly, it minimizes the distance between inter- classes which increases the false acceptance rates. This paper addresses these issues and proposes a robust algorithm that overcomes these limitations. The limitations are resolved through transforming pixels from non- illumination side to illuminated side. It has been revealed that proposed algorithm produced better results as compared to existing related algorithms.

A Nonlinear Hybrid Filter for Salt & Pepper Noise Removal from Color Images

Impulse noise reduction or removal is a very active research area of image processing. A nonlinear hybrid filter for removing fixed impulse noise (salt & pepper) noise from color images has been proposed in this study. Technique is based on mathematical morphology and trimmed standard median filter. Proposed filter is composed of a sequence of morphological standard and well known operations erosion-dilation and trimmed standard median filter. It removes the fixed impulse noise (salt & pepper) very well without distorting the image features, color components and edges. It does not introduce blurring and moving effects even in high noise densities (up to 90%). The standard similarity measure peak signal to noise ratio (PSNR) and computation time have been used to evaluate the performance of proposed hybrid filter.

Time signatures - an implementation of Keystroke and click patterns for practical and secure authentication

The analysis of keystroke dynamics (KD) is a developing biometric technique for user authentication. In computer security, its use is limited to some constraints such as longer typing and practice sessions. In this paper, a practical user authentication system is proposed that combines a conventional login/password method and a said biometric technique. The conventional password authentication method is enhanced through analysis of keystroke dynamics (KD) and click patterns (CP). In this way increased security level is achieved without using long and complicated passwords. For this, an application is developed to demonstrate the technique and the results are analyzed. user time signatures (TS) are identified after analyzing user KD and CP. Based on ability to follow their specific TS, users are categorized into beginner, standard and expert. At the time of login, the user inputs are matched with respective database records for authentication.

Appearance based pedestrians’ gender recognition by employing stacked auto encoders in deep learning

Abstract Pedestrians’ gender is a soft attribute which is useful in many areas of computer vision including human robot interaction, intelligent surveillance and human behavior analysis. Apart from its importance, pedestrians’ gender prediction is one of the challenging methodologies in image processing. In this article, a deep learning approach is presented to classify a pedestrian as a male or a female. As a pre-processing step, pedestrian parsing is performed by a deep decompositional neural network method. The outcome of this network is a binary mask that maps the pedestrian full body from the input image. The pedestrian body image is then extracted by applying the generated pedestrian mask to the input image. This pre-processed image is then supplied to the stacked sparse auto encoder with soft max classifier for prediction. The proposed network is trained and tested separately on different pedestrians’ views such as frontal views, back views and mixed views. The training is performed on PETA dataset. The experiments for testing are performed on MIT and PETA datasets (containing images other than train images). The accuracy values on MIT dataset are calculated as 82.9%, 81.8% and 82.4% on frontal, back and mixed views respectively. The mean AUC value by proposed scheme on PETA dataset is found as 91.5% ± 4. The performance measures and comparisons with existing works depict the robustness and applicability of proposed methodology.

An algorithm to find convex hull based on binary tree

In this paper, a new algorithm to find convex hull is presented which keeps the points into a form of nested binary tree. There are number of methods available for finding the convex hull but all of them are time consuming because they are based on comparing distances between the points and rotation of angles, whereas the proposed technique is based on binary tree which is efficient and more accurate than the existing algorithms as it decreases the computational cost and finds precise boundary of the object. Another main advantage of the proposed technique is the efficient memory usage because of binary tree structure.

Improving audio data quality and compression

High data quality at low bit rate is an essential goal that people want to achieve. It is necessary to transfer data at low bit rate so that the bandwidth of the medium can be utilized efficiently. In most of the speech coding techniques the goal of low bit rate transfer is achieved but the data quality is affected badly. The proposed technique is an attempt to improve the data quality at low bit rate as well as fast transmission of data. The proposed technique protects the data quality by applying Linear Predictive Coding-10 and achieves the low bit rate by applying Quadrature Mirror Filter. A comprehensive analysis is on the basis of given parameters as size, compression time, Signal to Noise Ratio, power, energy, power in air, energy in air, mean, standard deviation and intensity.

Fundus image classification methods for the detection of glaucoma: A review

Glaucoma is a neurodegenerative illness and is considered as a standout amongst the most widely recognized reasons for visual impairment. Nerves degeneration is an irretrievable procedure, so the diagnosis of the illness at an early stage is an absolute requirement to stay away from lasting loss of vision. Glaucoma effected mainly because of increased intraocular pressure, if it is not distinguished and looked early, it can result in visual impairment. There are not generally evident side effects of glaucoma; thus, patients attempt to get treatment just when the seriousness of malady is advanced altogether. Determination of glaucoma often comprises of review of the basic crumbling of the nerve in conjunction with the examination of visual function capacity. This article shows the persistent illustration of glaucoma, its side effects, and the potential people inclined to this malady. The essence of this article is on different classification methods being utilized and proposed by various scientists for the identification of glaucoma. This article audits a few division and segmentation methodologies that are exceptionally useful for recognizable proof, identification, and diagnosis of glaucoma. The research related to the findings and the treatment is likewise evaluated in this article.