Pixel Statistics Research Articles

Varying radix numeral system based adaptive image steganography

A novel image steganographic scheme based on the primitive and renowned numerical model is proposed. In this adaptive scheme the data to be embedded is dissected into numerals, each having variable information carrying capacity. The dissection is based on the statistics of the host image pixels i.e. the amount of adulteration that a pixel can tolerate. The proposed method provides proficient visual quality despite high payload capacity. The experimental results evaluated on 300 natural images show that the new scheme is resistant to RS steganalysis and offers high visual quality when compared with typical LSB-based schemes as well as their edge adaptive versions, like pixel-value-differencing-based schemes.

Local pixel statistics based impulse detection and hybrid color filtering for restoration of digital color images

This paper presents a two stage filtering system to remove random valued impulse noise from color images based on local statistics of the filtering window under consideration. In the first stage, to detect the noisy pixel, the locally adaptive threshold is derived from the pixels of the filtering window. In the second stage, the restoration of the noisy pixel is done on the basis of brightness and chromaticity information obtained from the neighbouring pixels in the filtering window. Simulation results show that the proposed scheme yields much superior performance in comparison with other color image filtering methods.

The Local Environments of Core-Collapse SNe within Host Galaxies

AbstractWe present constraints on core-collapse supernova progenitors through observations of their environments within host galaxies. This is achieved through 2 routes. Firstly, we investigate the spatial correlation of supernovae with host galaxy star formation using pixel statistics. We find that the main supernova types form a sequence of increasing association to star formation. The most logical interpretation is that this implies an increasing progenitor mass sequence going from the supernova type Ia arising from the lowest mass, through the type II, type Ib, and the supernova type Ic arising from the highest mass progenitors. We find the surprising result that the supernova type IIn show a lower association to star formation than type IIPs, implying lower mass progenitors. Secondly, we use host HII region spectroscopy to investigate differences in environment metallicity between different core-collapse types. We find that supernovae of types Ibc arise in slightly higher metallicity environments than type II events. However, this difference is not significant, implying that progenitor metallicity does not play a dominant role in deciding supernova type.

3D pose estimation and motion analysis of the articulated human hand-forearm limb in an industrial production environment

This study introduces an approach to model-based 3D pose estimation and instantaneous motion analysis of the human hand-forearm limb in the application context of safe human-robot interaction. 3D pose estimation is performed using two approaches: The Multiocular Contracting Curve Density (MOCCD) algorithm is a top-down technique based on pixel statistics around a contour model projected into the images from several cameras. The Iterative Closest Point (ICP) algorithm is a bottom-up approach which uses a motion-attributed 3D point cloud to estimate the object pose. Due to their orthogonal properties, a fusion of these algorithms is shown to be favorable. The fusion is performed by a weighted combination of the extracted pose parameters in an iterative manner. The analysis of object motion is based on the pose estimation result and the motion-attributed 3D points belonging to the hand-forearm limb using an extended constraint-line approach which does not rely on any temporal filtering. A further refinement is obtained using the Shape Flow algorithm, a temporal extension of the MOCCD approach, which estimates the temporal pose derivative based on the current and the two preceding images, corresponding to temporal filtering with a short response time of two or at most three frames. Combining the results of the two motion estimation stages provides information about the instantaneous motion properties of the object.Experimental investigations are performed on real-world image sequences displaying several test persons performing different working actions typically occurring in an industrial production scenario. In all example scenes, the background is cluttered, and the test persons wear various kinds of clothes. For evaluation, independently obtained ground truth data are used.

Fuzzy c-means clustering based on spatial neighborhood information for image segmentation

Fuzzy c-means (FCM) algorithm is one of the most popular methods for image segmentation. However, the standard FCM algorithm is sensitive to noise because of not taking into account the spatial information in the image. An improved FCM algorithm is proposed to improve the antinoise performance of FCM algorithm. The new algorithm is formulated by incorporating the spatial neighborhood information into the membership function for clustering. The distribution statistics of the neighborhood pixels and the prior probability are used to form a new membership function. It is not only effective to remove the noise spots but also can reduce the misclassified pixels. Experimental results indicate that the proposed algorithm is more accurate and robust to noise than the standard FCM algorithm.

Reversible Data Hiding Based on Adaptive Modulation of Statistics Invertibility

This paper proposes an improved method of reversible data hiding with increased capacity. The conventional method determines whether to embed a data bit in an image block according to the statistics of pixels in that block. Some images have pixel statistics that are inadequate for data hiding, and seldom or never have data embedded in them. The proposed method modulates the statistics invertibility to overcome such disadvantages, and is also able to improve the quality of the image containing the hidden data using block-adaptive modulation. Simulationresults show the effectiveness of the proposed method.

Seabed Segmentation Using Optimized Statistics of Sonar Textures

In this paper, we propose and compare two supervised algorithms for the segmentation of textured sonar images, with respect to seafloor types. We characterize seafloors by a set of empirical distributions estimated on texture responses to a set of different filters. Moreover, we introduce a novel similarity measure between sonar textures in this feature space. Our similarity measure is defined as a weighted sum of Kullback-Leibler divergences between texture features. The weight setting is twofold. First, each filter is weighted according to its discrimination power: The computation of these weights are issued from a margin maximization criterion. Second, an additional weight, evaluated as an angular distance between the incidence angles of the compared texture samples, is considered to take into account sonar-image acquisition process that leads to a variability of the backscattered value and of the texture aspect with the incidence-angle range. A Bayesian framework is used in the first algorithm where the conditional likelihoods are expressed using the proposed similarity measure between local pixel statistics and the seafloor prototype statistics. The second method is based on a variational framework as the minimization of a region-based functional that involves the similarity between global-region texture-based statistics and the predefined prototypes.

Markov random field-based clustering applied to the segmentation of masses in digital mammograms

In this paper we propose a new pixel clustering model applied to the analysis of digital mammograms. The clustering represents here the first step in a more general method and aims at the creation of a concise data-set (clusters) for automatic detection and classification of masses, which are typically among the first symptoms analysed in early diagnosis of breast cancer. For the purpose of this work, a set of mammographic images has been employed, that are 12-bit gray level digital scans and as such, are inherently inhomogeneous and affected by the noise resulting from the film scanning. The image pixels are described only by their intensity (gray level), therefore, the available information is limited to one dimension. We propose a Markov random field (MRF)-based technique that is suitable for performing clustering in an environment which is described by poor or limited data. The proposed method is a statistical classification model, that labels the image pixels based on the description of their statistical and contextual information. Apart from evaluating the pixel statistics, that originate from the definition of the K-means clustering scheme, the model expands the analysis by the description of the spatial dependence between pixels and their labels (context), hence leading to the reduction of the inhomogeneity of the output. Moreover, we define a probabilistic description of the model, that is characterised by a remarkable simplicity, such that its realisation can be easily and efficiently implemented in any high- or low-level programming language, thus allowing it to be run on virtually any kind of platform. Finally, we evaluate the algorithm against the classical K-means clustering routine. We point out similarities between the two methods and, moreover, show the advantages and superiority of the MRF scheme.

Scene-based nonuniformity correction and enhancement: pixel statistics and subpixel motion

We propose a framework for scene-based nonuniformity correction (NUC) and nonuniformity correction and enhancement (NUCE) that is required for focal-plane array-like sensors to obtain clean and enhanced-quality images. The core of the proposed framework is a novel registration-based nonuniformity correction super-resolution (NUCSR) method that is bootstrapped by statistical scene-based NUC methods. Based on a comprehensive imaging model and an accurate parametric motion estimation, we are able to remove severe/structured nonuniformity and in the presence of subpixel motion to simultaneously improve image resolution. One important feature of our NUCSR method is the adoption of a parametric motion model that allows us to (1) handle many practical scenarios where parametric motions are present and (2) carry out perfect super-resolution in principle by exploring available subpixel motions. Experiments with real data demonstrate the efficiency of the proposed NUCE framework and the effectiveness of the NUCSR method.

WND-CHARM: Multi-purpose image classification using compound image transforms

We describe a multi-purpose image classifier that can be applied to a wide variety of image classification tasks without modifications or fine-tuning, and yet provide classification accuracy comparable to state-of-the-art task-specific image classifiers. The proposed image classifier first extracts a large set of 1025 image features including polynomial decompositions, high contrast features, pixel statistics, and textures. These features are computed on the raw image, transforms of the image, and transforms of transforms of the image. The feature values are then used to classify test images into a set of pre-defined image classes. This classifier was tested on several different problems including biological image classification and face recognition. Although we cannot make a claim of universality, our experimental results show that this classifier performs as well or better than classifiers developed specifically for these image classification tasks. Our classifier’s high performance on a variety of classification problems is attributed to (i) a large set of features extracted from images; and (ii) an effective feature selection and weighting algorithm sensitive to specific image classification problems. The algorithms are available for free download from http://www.openmicroscopy.org.

Facial feature extraction using complex dual-tree wavelet transform

In this paper, we propose a novel method for facial feature extraction using the directional multiresolution decomposition offered by the complex wavelet transform. The dual-tree implementation of complex wavelet transform offered by Selesnick is used (DT-DWT(S)) [I.W., Selesnick, R.G. Baraniuk, N.C. Kingsbury, The dual-tree complex wavelet transform, IEEE Signal Processing Magazine, 6, s.l., IEEE, November 2005, vol. 22, pp. 123–151.]. In the dual-tree implementation, two parallel discrete wavelet transform (DWT) with different lowpass and highpass filters in different scales are used. The linear combination of subbands generated by two parallel DWT is used to generate 6 different directional subbands with complex coefficients. A test statistic, which is derived with absolute value of complex coefficient, whose distribution matches very closely with the directional information in the 6 subbands of the DT-DWT(S) is derived and used for detecting facial feature edges. The use of the complex wavelet transform is motivated by the fact that it helps eliminate the effects of non-uniform illumination, and the directional information provided by the different subbands makes it possible to detect edge features with different directionalities in the corresponding image. Edge information of facial area is enhanced using multiresolution structure of DT-DWT(S). The proposed method also employs an adaptive skin colour model instead of a predefined skin colour statistic. The model is developed with a unimodal Gaussian distribution using the skin region which is extracted excluding the detected edge map obtained from the DT-DWT(S). By combining the edge information obtained by using DT-DWT(S) and the non-skin areas obtained from the pixel statistics, the facial features are extracted. The algorithm is tested over the well known Carnegie Mellon University (CMU) and Marks Weber face databases. The average detection rate of the proposed method using DT-DWT(S) provides up to 9.6% improvement over the same method using discrete wavelet transform (DWT).

IMPROVING OBJECT DETECTION PERFORMANCE WITH GENETIC PROGRAMMING

This paper describes three developments to improve object detection performance using genetic programming. The first investigates three feature sets, the second investigates a new fitness function, and the third introduces a two phase learning method using genetic programming. This approach is examined on three object detection problems of increasing difficulty and compared with a neural network approach. The two phase GP approach with the new fitness function and the local concentric circular region features achieved the best results. The results suggest that the concentric circular pixel statistics are more effective than the square features for these object detection problems. The fitness function with program size is more effective and more efficient than without for these object detection problems and the evolved genetic programs using this fitness function are much shorter and easier to interpret. The two phase GP approach is more effective and more efficient than the single stage GP approach, and also more effective than the neural network approach on these problems using the same set of features.

On shadow elimination after moving region segmentation based on different threshold selection strategies

A popular approach for detecting moving object regions in video sequences is the application of the background subtraction technique. According to this technique the background (reference) image is subtracted from the current image frame and the moving parts are detected by the selection of a suitable threshold. In this paper we present our work to discriminate the moving pixels of the generated difference images from the relatively stationary pixels through the use of three different threshold selection strategies, namely, (i) ‘3 σ edit rule’, (ii) rule utilizing the Hampel identifier, and (iii) rule based on an ad hoc selection of threshold. Further, after segmentation a method of classification, based on a moving shadow search technique, previously developed by the authors, has been applied to segregate the moving shadow region from the actual moving object. The speed-up achieved through the use of the three aforementioned techniques on the core moving shadow search process, compared to that where no such process has been applied, has been documented. The final outcomes of applying the shadow detection technique after segmenting using each of the threshold selection strategies, one at a time, on some indoor video sequences have been demonstrated and comparison of the methods made.

Robust Environmental Change Detection Using PTZ Camera via Spatial-Temporal Probabilistic Modeling

This paper proposes a novel procedure for detecting environmental changes by using a pan-tilt-zoom (PTZ) camera. Conventional approaches based on pixel space and stationary cameras need time-consuming image registration to yield pixel statistics. This work proposes an alternative approach to describe each scene with a Gaussian mixture model (GMM) via a spatial-temporal statistical method. Although details of the environment covered by the camera are lost, this model is efficient and robust in recognizing scene and detecting scene changes in the environment. Moreover, the threshold selection for separating different environmental changes is convenient by using the proposed framework. The effectiveness of the proposed method is demonstrated experimentally in an office environment.

Fire detection using statistical color model in video sequences

In this paper, we propose a real-time fire-detector that combines foreground object information with color pixel statistics of fire. Simple adaptive background model of the scene is generated by using three Gaussian distributions, where each distribution corresponds to the pixel statistics in the respective color channel. The foreground information is extracted by using adaptive background subtraction algorithm, and then verified by the statistical fire color model to determine whether the detected foreground object is a fire candidate or not. A generic fire color model is constructed by statistical analysis of the sample images containing fire pixels. The first contribution of the paper is the application of real-time adaptive background subtraction method that aids the segmentation of the fire candidate pixels from the background. The second contribution is the use of a generic statistical model for refined fire-pixel classification. The two processes are combined to form the fire detection system and applied for the detection of fire in the consecutive frames of video sequences. The frame-processing rate of the detector is about 40 fps with image size of 176 × 144 pixels, and the algorithm’s correct detection rate is 98.89%.

An adaptive squared-distance-weighted interpolation for volume reconstruction in 3D freehand ultrasound

Volume reconstruction is a key procedure in 3D ultrasound imaging. An algorithm named as squared-distance-weighted (SDW) interpolation has been earlier proposed to reduce the blurring effect in the 3D ultrasonic images caused by the conventional distance weighted (DW) interpolation. However, the SDW parameter α, which controls the weight distribution, is a constant assigned by operators so that the interpolation effect is invariant for both sharp edges and speckle noises. In this paper, we introduced a new adaptive algorithm based on SDW interpolation for volume reconstruction of 3D freehand ultrasound. In the algorithm, the local statistics of pixels surrounding each voxel grid were used to adaptively adjust the parameter α in SDW. The voxel grids with a higher ratio of local variance and mean in their neighbourhoods would have a smaller α to make the image details sharper, while the voxel grids locating in regions with a lower ratio of local variance and mean would have a larger α to smooth image content in homogeneous regions, where speckle noise is usually observed and damages the image quality. By comparing the simulation results using the SDW and new adaptive algorithm, it was demonstrated that this new algorithm worked well in both edge preservation and speckle reduction.

Correction of phase offset errors in main pulmonary artery flow quantification

To investigate whether an existing method for correction of phase offset errors in phase-contrast velocity quantification is applicable for assessment of main pulmonary artery flow with an MR scanner equipped with a high-power gradient system. The correction method consists of fitting a surface through the time average of stationary pixels of velocity-encoded phase images, and subtracting this surface from the velocity images. Pixels are regarded as stationary if their time standard deviation falls into the lowest percentile. Flow was measured in the main pulmonary artery of 15 subjects. Each measurement was repeated on a stationary phantom. The phase offset error in the phantom was used as a reference. Correction was applied with varying polynomial surface orders (0-5) and stationarity percentiles (5-50%). The optimal surface order and stationarity percentile were determined by comparing the fitted surface with the phantom. Using a first-order surface and a (noncritical) 25% percentile, the correction method significantly reduced the phase offset error from 1.1 to 0.35 cm/second (RMS), which is equivalent to a reduction from 11% to 3.3% of mean volume flow. Phase error correction strongly affected stroke volume (range -11 to 26%). The method significantly reduces phase offset errors in pulmonary artery flow.

A Bayesian MRF framework for labeling terrain using hyperspectral imaging

Studies of hyperspectral images point to non-Gaussian statistics of pixels values, and consequently, standard Gaussian models may not perform well in hyperspectral image analysis. This paper presents novel probability models that capture non-Gaussian statistics of hyperspectral images, and uses them in automated classification of terrain sites. After the data are preprocessed using standard dimension-reduction tools, we use: 1) a nonparametric density estimate for capturing spectral variation at each site and 2) two parametric families-generalized Laplacian and Bessel K form-to capture non-Gaussian statistics of difference pixels. Assuming an Ising-type prior on site labels, favoring a smooth classification, we formulate a Markov random field-maximum a posteriori estimation problem and use a Markov chain to estimate site classifications. Results are presented from application of this framework to Washington, DC Mall and Indian Springs rural area datasets.

Interaction of luminance and higher-order statistics in texture discrimination

Most studies of texture processing are based on textures in which individual pixel statistics are varied and spatial correlations are absent (“IID textures”), or textures in which spatial correlation structure is varied and luminance, or first-order, statistics are held constant. Here we jointly examine simple pixel statistics and fourth-order spatial correlation structure along the continuum of “even” and “odd” isodipole textures of Julesz, Gilbert and Victor [Julesz, B., Gilbert, E.N., & Victor, J.D. (1978). Biological Cybernetics, 31(3) 137–140], as well as their interactions. Absolute efficiency to detect either kind of statistical cue is low: approximately 0.05 for luminance statistics, and 0.004 for isodipole statistics. Above threshold, isodipole statistics must change by approximately four times the amount that pixel statistics must change to generate an equally salient texture. When pixel statistics and isodipole statistics are simultaneously varied, the two texture cues combine by probability summation and perceptual distances are approximately Euclidean. Superimposed on this picture are subtle foreground/background asymmetries that suggest properties of the visual mechanisms that are sensitive to these image statistics.

Metallicity of the Intergalactic Medium Using Pixel Statistics. II. The Distribution of Metals as Traced by Civ

We measure the distribution of carbon in the intergalactic medium as a function of redshift z and overdensity δ. Using a hydrodynamical simulation to link the H I absorption to the density and temperature of the absorbing gas, and a model for the UV background radiation, we convert ratios of C IV to H I pixel optical depths into carbon abundances. For the median metallicity this technique was described and tested in Paper I of this series. Here we generalize it to reconstruct the full probability distribution of the carbon abundance and apply it to 19 high-quality quasar absorption spectra. We find that the carbon abundance is spatially highly inhomogeneous and is well described by a lognormal distribution for fixed δ and z. Using data in the range log δ = -0.5-1.8 and z = 1.8-4.1, and a renormalized version of the 2001 Haardt & Madau model for the UV background radiation from galaxies and quasars, we measure a median metallicity of [C/H] = -3.47 + 0.08(z - 3) + 0.65(log δ - 0.5) and a lognormal scatter of σ([C/H]) = 0.76 + 0.02(z - 3) - 0.23(log δ - 0.5). Thus, we find significant trends with overdensity but no evidence for evolution. These measurements imply that gas in this density range accounts for a cosmic carbon abundance of [C/H] = -2.80 ± 0.13 (ΩC ≈ 2 × 10-7), with no evidence for evolution. The dominant source of systematic error is the spectral shape of the UV background, with harder spectra yielding higher carbon abundances. While the systematic errors due to uncertainties in the spectral hardness may exceed the quoted statistical errors for δ < 10, we stress that UV backgrounds that differ significantly from our fiducial model give unphysical results. The measured lognormal scatter is strictly independent of the spectral shape, provided the background radiation is uniform. We also present measurements of the C III/C IV ratio (which rule out temperatures high enough for collisional ionization to be important for the observed C IV) and of the evolution of the effective Lyα optical depth.