Difference Detection Between Two Images for Image Monitoring

In manufacturing industries, images are commonly used for quality control purposes. In such applications, if the quality of the products is good, then their images should be all similar to the image of a good-quality product. Therefore, comparison of images is a fundamental task in image-based quality control. This problem, however, is complicated in the sense that (1) observed images often contain noise, and (2) the related images need to be geometrically matched up first because images of different products could be geometrically mismatched because the relative positions between a camera and different products are often not exactly the same. The first issue requires a statistical method that can remove noise, and the second issue is related to the so-called image registration problem in the image processing literature. In this article, we propose effective methods for detecting difference between two images of products, and our proposed methods can accommodate both noise and geometric mismatch mentioned above. Theoretical results and numerical examples show that they can work effectively in applications.     

Pixel expansion-free grey-scale image sharing method

Abstract

In recent years, many visual secret sharing technologies have been proposed to protect the security of secret images (black and white, grey scale or colour images). In 2005, Lukac and Plataniotis used the concept of the Naor–Shamir method to propose a visual secret sharing technology for sharing secret grey-scale images based on bit-plane decomposition. Although their method can avoid pixel-value cutting problem, it still suffers from the pixel expansion problem. In this paper, we propose a new secret grey-scale image sharing method to improve this situation. In the proposed sharing image creation phase, two sharing images (one is grey-scale and the other is binary) are created and later shared by two protectors. The original secret image can be easily reconstructed if both sharing images are obtained. However, one cannot obtain the original secret image from each of the sharing images. Experimental results also show that the proposed method can effectively solve the pixel expansion problem.     

Detecting aircrafts from satellite images using saliency and conical pyramid based template representation

Automatic target localization in satellite images still remains as a challenging problem in the field of computer vision. The issues involved in locating targets in satellite images are viewpoint, spectral (intensity) and scale variations. Diversity in background texture and target clutter also adds up to the complexity of the problem of localizing aircrafts in satellite images. Failure of modern feature extraction and object detection methods highlight the complexity of the problem. In the proposed work, pre-processing techniques, viz. denoising and contrast enhancement, are first used to improve the quality of the images. Then, the concept of unsupervised saliency is used to detect the potential regions of interest, which reduces the search space. Parts from the salient regions are further processed using clustering and morphological processing to get the probable regions of isolated aircraft targets. Finally, a novel conical pyramid based framework for template representation of the target samples is proposed for matching. Experimental results shown on a few satellite images exhibit the superior performance of the proposed methods.     

NOSER: An algorithm for solving the inverse conductivity problem

The inverse conductivity problem is the mathematical problem that must be solved in order for electrical impedance tomography systems to be able to make images. Here we show how this inverse conductivity problem is related to a number of other inverse problem. We then explain the workings of an algorithm that we have used to make images from electrical impedance data measured on the boundary of a circle in two dimensions. This algorithm is based on the method of least squares. It takes one step of a Newton's method, using a constant conductivity as an initial guess. Most of the calculations can therefore be done analytically. The resulting code is named NOSER, for Newton's One-Step Error Reconstructor. It provides a reconstruction with 496 degrees of freedom. The code does not reproduce the conductivity accurately (unless it differs very little from a constant), but it yields useful images. This is illustrated by images reconstructed from numerical and experimental data, including data from a human chest.     

High-Resolution Digital Integral Photography by use of a Scanning Microlens Array

We suggest what we believe is a new three-dimensional (3-D) camera system for integral photography. Our method enables high-resolution 3-D imaging. In contrast to conventional integral photography, a moving microlens array (MLA) and a low-resolution camera are used. The intensity distribution in the MLA image plane is sampled sequentially by use of a pinhole array. The inversion problem from pseudoscopic to orthoscopic images is dealt with by electronic means. The new method is suitable for real-time 3-D imaging. We verified the new method experimentally. Integral photographs with a resolution of 3760 pixels x 2560 pixels (188 x 128 element images) are presented.     

Topographic profiling and refractive-index analysis by use of differential interference contrast with bright-field intensity and atomic force imaging

A methodology is described for phase restoration of an object function from differential interference contrast (DIC) images. The methodology involves collecting a set of DIC images in the same plane with different bias retardation between the two illuminating light components produced by a Wollaston prism. These images, together with one conventional bright-field image, allows for reduction of the phase deconvolution restoration problem from a highly complex nonlinear mathematical formulation to a set of linear equations that can be applied to resolve the phase for images with a relatively large number of pixels. Additionally, under certain conditions, an on-line atomic force imaging system that does not interfere with the standard DIC illumination modes resolves uncertainties in large topographical variations that generally lead to a basic problem in DIC imaging, i.e., phase unwrapping. Furthermore, the availability of confocal detection allows for a three-dimensional reconstruction with high accuracy of the refractive-index measurement of the object that is to be imaged. This has been applied to reconstruction of the refractive index of an arrayed waveguide in a region in which a defect in the sample is present. The results of this paper highlight the synergism of far-field microscopies integrated with scanned probe microscopies and restoration algorithms for phase reconstruction.     

Improved chicken swarm optimization to classify dementia MRI images using a novel controlled randomness optimization algorithm

The objective of this research paper is to categorize the magnetic resonance imaging (MRI) images as demented (DEM) or nondemented (ND) using improved chicken swarm optimization technique (ICSO). In literature, CSO technique is widely used to solve numerical optimization and feature selection problem. Using this optimization technique for medical image classification problem will be a pioneering idea. If this technique is directly used to classify the medical images, it provides poor results. Hence, appropriate enhancements are made on the original algorithm using a novel controlled randomness optimization algorithm and control parameter tuning. Cross-over and Rooster selection methods are also implemented in cascaded manner for further performance improvization. All the experiments are made for two cases: with and without statistical features. The brain MRI images of 65 ND and 52 DEM subjects obtained from the Open Access Series of Imaging Studies website are used in this analysis. The ICSO without statistical features provides the highest accuracy of 86.32%, whereas the original chicken swarm optimization technique provides the accuracy of 52.13% and 52.99% with and without statistical features, respectively.     

Measuring Statistical Geometric Properties of Tomographic Images of Soils

Porous media modeling is relevant in several applications, such as agricultural engineering, where soil compaction analysis requires the estimation of soil transport properties. For example, the prediction of root growing patterns and their environmental impact is usually measured by analyzing soil fluid infiltration capacity and water retention. Recently, tomographic images have been used in nondestructive tests of soil. However, using such images is challenging for two reasons: (1) Tomographic images are usually noisy, which complicates their segmentation, and (2) modeling the soil structure requires establishing adjacency relations among neighboring tomographic slices, which has a significant computational cost due to the combinatorial nature of this problem. In this paper, we propose a solution for both problems. The experimental results show that soil samples can be analyzed and classified with significant accuracy using our proposed approach.     

Improved cuckoo search with particle swarm optimization for classification of compressed images

The need for a general purpose Content Based Image Retrieval (CBIR) system for huge image databases has attracted information-technology researchers and institutions for CBIR techniques development. These techniques include image feature extraction, segmentation, feature mapping, representation, semantics, indexing and storage, image similarity-distance measurement and retrieval making CBIR system development a challenge. Since medical images are large in size running to megabits of data they are compressed to reduce their size for storage and transmission. This paper investigates medical image retrieval problem for compressed images. An improved image classification algorithm for CBIR is proposed. In the proposed method, RAW images are compressed using Haar wavelet. Features are extracted using Gabor filter and Sobel edge detector. The extracted features are classified using Partial Recurrent Neural Network (PRNN). Since training parameters in Neural Network are NP hard, a hybrid Particle Swarm Optimization (PSO) – Cuckoo Search algorithm (CS) is proposed to optimize the learning rate of the neural network.     

A 100-MHz ultrasound imaging system for dermatologic and ophthalmologic diagnostics

A major design problem concerning high-frequency broad-band ultrasound imaging systems is caused by the strong dispersive attenuation of the tissue, which gives rise to images with inhomogeneous resolution and poor signal to noise ratio (SNR). To address the noise problem, strongly focused transducers with high energy density in a narrow focal region are utilized, which also provide more isotropic images due to improved lateral resolution. To account for the short depth of the focal area two suitable imaging conceptions are used: 1) synthetic aperture concept and 2) B/D-scan concept. To avoid the inhomogeneity of the images, different transmitter signals for each depth are applied, which are pseudoinversely prefiltered according to the transfer function of the tissue. To gain signal energy required for inverse filtering, a pulse compression technique with nonlinearly frequency modulated chirp signals is utilized. These procedures have been implemented in an ultrasound imaging system, which has been developed in the authors' laboratory for eye and skin examinations, It can be used with transducers in a frequency range from 20 to 250 MHz.     

Wavelet restoration of medical pulse-echo ultrasound images in an EM framework

The clinical utility of pulse-echo ultrasound images is severely limited by inherent poor resolution that impacts negatively on their diagnostic potential. Research into the enhancement of image quality has mostly been concentrated in the areas of blind image restoration and speckle removal, with little regard for accurate modeling of the underlying tissue reflectivity that is imaged. The acoustic response of soft biological tissues has statistics that differ substantially from the natural images considered in mainstream image processing: although, on a macroscopic scale, the overall tissue echogenicity does behave some-what like a natural image and varies piecewise-smoothly, on a microscopic scale, the tissue reflectivity exhibits a pseudo-random texture (manifested in the amplitude image as speckle) due to the dense concentrations of small, weakly scattering particles. Recognizing that this pseudorandom texture is diagnostically important for tissue identification, we propose modeling tissue reflectivity as the product of a piecewise-smooth echogenicity map and a field of uncorrelated, identically distributed random variables. We demonstrate how this model of tissue reflectivity can be exploited in an expectation-maximization (EM) algorithm that simultaneously solves the image restoration problem and the speckle removal problem by iteratively alternating between Wiener filtering (to solve for the tissue reflectivity) and wavelet-based denoising (to solve for the echogenicity map). Our simulation and in vitro results indicate that our EM algorithm is capable of producing restored images that have better image quality and greater fidelity to the true tissue reflectivity than other restoration techniques based on simpler regularizing constraints.     

Enhancing CNN for Forensics Age Estimation Using CGAN and Pseudo-Labelling

Age estimation using forensics odontology is an important process in identifying victims in criminal or mass disaster cases. Traditionally, this process is done manually by human expert. However, the speed and accuracy may vary depending on the expertise level of the human expert and other human factors such as level of fatigue and attentiveness. To improve the recognition speed and consistency, researchers have proposed automated age estimation using deep learning techniques such as Convolutional Neural Network (CNN). CNN requires many training images to obtain high percentage of recognition accuracy. Unfortunately, it is very difficult to get large number of samples of dental images for training the CNN due to the need to comply to privacy acts. A promising solution to this problem is a technique called Generative Adversarial Network (GAN). GAN is a technique that can generate synthetic images that has similar statistics as the training set. A variation of GAN called Conditional GAN (CGAN) enables the generation of the synthetic images to be controlled more precisely such that only the specified type of images will be generated. This paper proposes a CGAN for generating new dental images to increase the number of images available for training a CNN model to perform age estimation. We also propose a pseudo-labelling technique to label the generated images with proper age and gender. We used the combination of real and generated images to train Dental Age and Sex Net (DASNET), which is a CNN model for dental age estimation. Based on the experiment conducted, the accuracy, coefficient of determination (R2) and Absolute Error (AE) of DASNET have improved to 87%, 0.85 and 1.18 years respectively as opposed to 74%, 0.72 and 3.45 years when DASNET is trained using real, but smaller number of images.     

MAGNet融合导向滤波的真实图像去雾方法

为了提高单幅图像去雾方法的准确性以及适用范围,提出一种MAGNet融合导向滤波的真实图像去雾方法.首先,根据雾在真实图像中分布特性以及成像原理,设计多注意力残差密集块,从而有效提取真实图像中与雾相关的特征并降低梯度消失风险;其次,构造基于所设计的多注意力残差密集块的端到端卷积神经网络,实现对有雾图像中雾的去除;最后,将...     

Scanline-based stereo matching by genetic algorithms

The computer stereo-vision system (CSVS) is a non-emissive, non-scanning, economic and flexible vision system. It adopts a pair of cameras at different positions and acquires simultaneously two images of an object to derive information on depth. The problem of correspondence between images is one of most important issues in the study of CSYS. The key concern of this problem is to determine which feature in one image corresponds to a given feature in the other image. Owing to the high complexity of the correspondence problem, the matching process usually suffers from slow computation time and imprecise results. The present paper proposes a scanline-~ased stereo-matching model. The objective function of the proposed model is a function with the means and variance of intensity, gradient magnitude and direction, and related constraints including uniqueness, geometry and epipolar constraints. The formulated energy function is then solved using genetic algorithms. Binary representation is used to match the relationship between the candidate features of both images. Two operators, the roulette selection operator and elite principle, are adopted for offspring selection. A modified position-based crossover operator and its repair mechanism are developed to exchange partially the chromosomes. In addition, an order-based mutation is used to prevent early convergence. Experiments show that the designed selection, crossover and mutation operators can effectively derive the matching relations line by line for real-world testing images. The proposed method was also compared with another method and was superior in computation speed and required less demand of the memory.     

Automated segmentation of synchrotron radiation micro-computed tomography biomedical images using Graph Cuts and neural networks

Synchrotron Radiation (SR) X-ray micro-Computed Tomography (μCT) enables magnified images to be used as a non-invasive and non-destructive technique with a high space resolution for the qualitative and quantitative analyses of biomedical samples. The research on applications of segmentation algorithms to SR-μCT is an open problem, due to the interesting and well-known characteristics of SR images for visualization, such as the high resolution and the phase contrast effect. In this article, we describe and assess the application of the Energy Minimization via Graph Cuts (EMvGC) algorithm for the segmentation of SR-μCT biomedical images acquired at the Synchrotron Radiation for MEdical Physics (SYRMEP) beam line at the Elettra Laboratory (Trieste, Italy). We also propose a method using EMvGC with Artificial Neural Networks (EMANNs) for correcting misclassifications due to intensity variation of phase contrast, which are important effects and sometimes indispensable in certain biomedical applications, although they impair the segmentation provided by conventional techniques. Results demonstrate considerable success in the segmentation of SR-μCT biomedical images, with average Dice Similarity Coefficient 99.88% for bony tissue in Wistar Rats rib samples (EMvGC), as well as 98.95% and 98.02% for scans of Rhodnius prolixus insect samples (Chagas's disease vector) with EMANNs, in relation to manual segmentation. The techniques EMvGC and EMANNs cope with the task of performing segmentation in images with the intensity variation due to phase contrast effects, presenting a superior performance in comparison to conventional segmentation techniques based on thresholding and linear/nonlinear image filtering, which is also discussed in the present article.     

Multi-feature based automatic recognition of ship targets in ISAR

This paper deals with the problem of non-cooperative target recognition. Specifically, the aim is the automatic recognition of ship targets from inverse synthetic aperture radar (ISAR) images. For this purpose a new two-step multi-feature based technique is proposed; this technique uses a number of features extracted from the ship radar image and matches these features with those extracted from the images obtained by properly projecting the target models of the classification library. Both cases of a priori known or unknown ship aspect angles are considered: the knowledge of the ship aspect (as available from tracking data) allows the selection of the candidate models on the basis of the matching between the ship and the model length, thus resulting in a performance improvement. Moreover, both single- and multi-frame-based processing techniques are proposed in order to assess the performance improvement achievable when an increasing number of ISAR images are involved in the decision; the fusion strategy adopted for the exploitation of the information from the multiple images is also described. The performance of the overall proposed technique is deeply investigated against simulated data. Results of its application to a set of live ISAR images of a ship target are also provided showing the effectiveness of the proposed approach.     

一种有效的机场安检X光手提行李图像两级增强方法

针对低对比度X光手提行李图像在机场安检中容易产生高虚警或高漏警的问题,提出了一种两级X光图像增强方法.首先,应用离散小波变换和独立分量分析方法对低能和高能X光图像去噪并融合,实现一级增强.然后,利用本文提出的自适应正弦灰度变换实现二级增强.实验结果表明,本文方法能够有效地改善图像质量,优于文中给出的其它增强方法,更有利...     

Performance evaluation of two-dimensional phase unwrapping algorithms

We present a performance evaluation of eight two-dimensional phase unwrapping methods with respect to correct phase unwrapping and execution times. The evaluated methods are block least squares (BLS), adaptive integration (AI), quality guided path following (QUAL), mask cut (MCUT), multigrid (MGRID), preconditioned conjugate gradient (PCG), Flynn's (FLYNN), and Liang's (LIANG). This set included integration- (path following), least-squares-, L(1)-, and model-based methods. The methods were tested on several synthetic images, on two magnetic resonance images, and on two interferometry images. The synthetic images were designed to demonstrate different aspects of the phase unwrapping problem. To test the noise robustness of the methods, independent noise was added to the synthetic images to yield different signal-to-noise ratios. Each experiment was performed 50 times with different noise realizations to test the stability of the methods. The results of the experiments showed that the congruent minimum L(1) norm FLYNN method was best overall and the most noise robust of the methods, but it was also one of the slowest methods. The integration-based QUAL method was the only method that correctly unwrapped the two interferometry images. The least-squares-based methods (MGRID, PCG) gave worse results on average than did the integration- (or path following) based methods (BLS, AI, QUAL, MCUT) and were also slower. The model-based LIANG method was sensitive to noise and resulted in large errors for the magnetic resonance images and the interferometry images. In conclusion, for a particular application there is a trade-off between the quality of the unwrapping and the execution time when we attempt to select the most appropriate method.     

机车全动态包络线测量中的目标点自动识别算法研究

针对机车动态限界图像中目标点自动识别的问题,提出一种基于双分辨力分析的目标点识别算法,解决了高分辨力图像对于多目标的搜索效率问题,并提出了一种自描述向量用于改进传统的RANSAC算法,以减少背景噪声的干扰,完成了不同限界图像目标点的单应矩阵的求解问题。最终实现高效且准确的目标点自动识别与匹配。     

Microlens arrays for integral imaging system

When designing a system capable of capturing and displaying 3D moving images in real time by the integral imaging (II) method, one challenge is to eliminate pseudoscopic images. To overcome this problem, we propose a simple system with an array of three convex lenses. First, the lateral magnification of the elemental optics and the expansion of an elemental image is described by geometrical optics, confirming that the elemental optics satisfies the conditions under which pseudoscopic images can be avoided. In using the II method, adjacent elemental images must not overlap, a condition also satisfied by the proposed optical system. Next, an experiment carried out to acquire and display 3D images is described. The real-time system we have constructed comprises an elemental optics array with 54 H x 59 V elements, a CCD camera to capture a group of elemental images created by the lens array, and a liquid crystal panel to display these images. The results of the experiment confirm that the system produces orthoscopic images in real time, and thus is effective for real-time application of the II method.