Niladri B. Puhan

Efficient segmentation technique for noisy frontal view iris images using Fourier spectral density

In a less constrained capture of iris images to build a high-speed iris recognition system, the design of a robust and fast iris segmentation method is important. In this paper, a new iris segmentation technique based on the Fourier spectral density is proposed for noisy frontal view eye images captured with minimum cooperation from the subjects. The proposed segmentation method is not an iterative technique and it runs in deterministic time. The computational complexity of the proposed method is found to be significantly lower than the existing approaches based on integro-differential operator, Hough transform and active contour. The basic idea underlying the proposed method is to localize the limbic and pupil boundaries using the Fourier spectral density. The performance studies on a recently created iris database, called UBIRIS (Proenca and Alexandre in Lect Notes Comput Sci 3617:970–977, 2005) containing defocused, reflection-contained and eyelid-occluded iris images in visible spectral range, show that the proposed method is much faster than the existing methods and simultaneously achieves good segmentation accuracy.

A novel iris database indexing method using the iris color

In a very large iris database, the design of an effective iris indexing method reduces the computational complexity of feature matching process. This paper presents a new iris indexing technique based on the iris color for noisy iris images. The proposed method using the chrominance components achieves high indexing performance in defocused, reflection-contained and eyelid-occluded iris images. The performance measures such as the hit rate and the penetration rate computed on a recently created noisy iris database, namely UBIRIS (Proenca and Alexandre, 2005), show the effectiveness of the iris color to index very large iris databases.

Robust eyeball segmentation in noisy iris images using fourier spectral density

In this paper, a new eyeball segmentation approach based on the Fourier spectral density is proposed for noisy iris images. The design of an accurate segmentation method for noisy iris images could make non-cooperative iris recognition possible. The proposed segmentation method aims to achieve high segmentation accuracy in defocused, reflection-contained and eyelid-occluded iris images. The proposed method could extract the eyeball region correctly in a significant number of noisy iris images from the UBIRIS database [16].

Binary document image watermarking for secure authentication using perceptual modeling

In this paper, we propose a new perception based watermarking algorithm in binary document images for secure authentication purpose. Binary image watermarking with pixel flipping approach is a challenging problem, because flipping the black and white pixels in such simple images can bring noticeable visual distortion. A novel perceptual based model was proposed towards digital watermarking of binary images in the paper of A.T.S. Ho et al. (May 2004). The model estimates the distortion resulting from flipping of a pixel by finding the curvature-weighted distance difference (CWDD) measure between original and watermarked contour segments. In this paper, the reversible property of the CWDD measure is used towards designing a new authentication watermarking algorithm so that the possibility of any undetected modification to the watermarked image is removed. This algorithm embeds an authentication signature computed from the original image into itself after identifying an ordered set of low-distortion pixels. The same ordered set of pixels are correctly found in both the embedder and blind detector through the design of necessary conditions. The ability of the proposed authentication algorithm to detect any modification in the watermarked image is equivalent to the security of cryptographic authentication. The parity attack found in the previous block-wise data hiding methods in binary images is not possible in the proposed algorithm due to pixel-wise embedding of the authentication signature. Simulation results show the imperceptibility of the watermarking process and successful detection of content modifications

A new iris liveness detection method against contact lens spoofing

Liveness detection is an important and challenging issue in iris recognition system security against spoofing. In this paper, a new iris liveness detection method is proposed against semi-transparent contact lens based spoofing. The proposed method works through iris texture dissimilarity between two iris regions due to pupillary light reflex. The texture dissimilarity is computed in the iris region that can be superimposed by a contact lens and it exists from the expanded pupil to outer iris boundary. The normalized Hamming distance computed on binary texture features is used to quantify textural dissimilarity between localized iris regions. Simulation results show that live iris textures produce high values of Hamming distance with the decreasing trend from inner to outer iris boundary.

Perception based binary image watermarking

The use of a suitable perceptual model is necessary to minimize visual distortion in marked images, because minor modifications to the pixels can be perceptible since the pixels are either black or white. In this paper, a new perceptual model is proposed for binary images that is useful for data hiding applications. In our model, the distortion that occurs after flipping a pixel is estimated on the novel curvature-weighted distance difference (CWDD) measure between two contour segments. Through subjective tests the perceptual measure is validated and highly correlated with human perception.

Off-line signature verification: upper and lower envelope shape analysis using chord moments

Signature is an important and useful behavioural biometric which exhibits significant amount of non-linear variability. In this study, the authors concentrate on finding an envelope shape feature known as ‘chord moments’. Central moments such as the variance, skewness and kurtosis along with the first moment (mean) are computed from sets of chord lengths and angles for each envelope reference point. The proposed chord moments adequately quantify the spatial inter-relationship among upper and lower envelope points. The moment-based approach significantly reduces the dimension of highly detailed chord sets and is experimentally found to be robust in handling non-linear variability from signature images. The proposed chord moments coupled with the support vector machine classifier lead to a writer dependent off-line signature verification system that achieves state-of-the-art performance on the noisy Center of Excellence for Document Analysis and Recognition database.

Handwritten numeral recognition using non-redundant Stockwell transform and bio-inspired optimal zoning

Handwritten digit recognition is one of the challenging problems of character recognition because of the large variation in writing styles of individuals and the presence of similar looking shapes of different numerals. Most of the feature extraction techniques are based on statistical or topological attributes of the image in its spatial domain, barring few works attempting feature extraction in a transformed domain. Another challenge is the optimal selection of zones while extracting features from localised zones of the unknown (test) image. In most of the cases, the recognition phase, being isolated from the training phase makes it impossible to adaptively improve the feature selection using the knowledge obtained from error analysis. In this study, the authors propose a feature extraction technique, new to the character recognition problem, using non-redundant Stockwell transform. Another transformed domain feature extraction using Slantlet coefficients is proposed. They also propose to use bio-inspired and evolutionary computing-based optimisation techniques to adaptively select the optimal zone arrangement in the feature selection stage from the knowledge of classification accuracy. The proposed methods are experimentally validated on handwritten digit database of Odia language which proves to outperform any recognition accuracy reported before.

Secure authentication watermarking for localization against the Holliman---Memon attack

Authentication watermarking schemes using block-wise watermarks for tamper localization are vulnerable to the Holliman–Memon attack. In this paper, we propose a novel method based on the Wong’s localization scheme (Proceedings of the IS&T PIC, Portland) to resist this attack. A unique image index scheme is used for computing the authentication signature that is embedded in the least significant bit-plane of the block. The informed detector estimates the correct image index by using the side information about the watermarked image. The image index estimation from the fake image can definitely be an alternative to keeping a directory of image indices. So it is not necessary to manage the database of image indices for the verification purpose. The authenticity measure is defined to quantify the attack severity by taking the connectivity among possible authentic blocks into consideration. There are more blocks verified as authentic when this measure is high for a fake image constructed using this attack. As such, the blocks for a fake image can be chosen from a reduced number of database images. The blocks from any such image are to be connected with each other to maximize the authenticity measure. Thus, the attacker’s task to generate a fake image of reasonable perceptual quality becomes increasingly difficult. With the proposed method there is no loss or ambiguity in localization after the Holliman–Memon attack and content tampering in an image. The localization accuracy in the proposed method is demonstrated by the simulation results and is equal to the chosen block size, similar to the Wong’s scheme.

Erasable Authentication Watermarking in Binary Document Images

In this paper, we propose a new secure authentication method in binary document images using erasable watermarks. For localization, a sufficient number of low-distortion pixels may not be available in a block and to embed the authentication signature with blind detection constraint poses a challenging problem. Also, a perceptual watermark cannot be embedded in white or background regions of the document image, making such regions insecure against hostile attacks. In the proposed method, an erasable watermark is designed and embedded in each block of a document image for localization against tampering and Holliman-Memon attack. After verifying the content of each block, the exact copy of original image can be restored at the blind detector.