BE-ACGAN: Photo-realistic residual bit-depth enhancement by advanced conditional GAN

Since more demands for high quality visualization have been raised in various fields, monitors with higher bit-depth (HBD) become popular in recent years. However, most digital images are at low bit-depth (LBD) and usually of low visual quality with annoying false contours when displayed on HBD monitors directly. To reconstruct visually pleasant HBD images, many bit-depth enhancement (BE) algorithms have been proposed from various aspects, but the recovered HBD images are usually unsatisfactory with conspicuous false contours or over-blurred textures. Inspired by discriminative learning, we propose a residual BE algorithm based on advanced conditional generative adversarial network (BE-ACGAN), in which the discriminator adversarially helps assess image quality and train the generator to achieve more photo-realistic recovery performance. Besides, since it is hard to distinguish between the reconstructed and real HBD images with similar structures, the discriminator takes residual images as input and further takes LBD images as conditions to achieve more reliable performance. In addition, we present a novel loss function to deal with the difficulty of unstable adversarial training. The proposed algorithm outperforms the state-of-the-art methods on large-scale benchmark datasets. Source codes are available at https://github.com/TJUMMG/BE- ACGAN/.     

Retinal Vessel Extraction Framework Using Modified Adaboost Extreme Learning Machine

An explicit extraction of the retinal vessel is a standout amongst the most significant errands in the field of medical imaging to analyze both the ophthalmological infections, for example, Glaucoma, Diabetic Retinopathy (DR), Retinopathy of Prematurity (ROP), Age-Related Macular Degeneration (AMD) as well as non retinal sickness such as stroke, hypertension and cardiovascular diseases. The state of the retinal vasculature is a significant indicative element in the field of ophthalmology. Retinal vessel extraction in fundus imaging is a difficult task because of varying size vessels, moderately low distinction, and presence of pathologies such as hemorrhages, microaneurysms etc. Manual vessel extraction is a challenging task due to the complicated nature of the retinal vessel structure, which also needs strong skill set and training. In this paper, a supervised technique for blood vessel extraction in retinal images using Modified Adaboost Extreme Learning Machine (MAD-ELM) is proposed. Firstly, the fundus image preprocessing is done for contrast enhancement and in-homogeneity correction. Then, a set of core features is extracted, and the best features are selected using “minimal Redundancy-maximum Relevance (mRmR).” Later, using MAD-ELM method vessels and non vessels are classified. DRIVE and DR-HAGIS datasets are used for the evaluation of the proposed method. The algorithm’s performance is assessed based on accuracy, sensitivity and specificity. The proposed technique attains accuracy of 0.9619 on the DRIVE database and 0.9519 on DR-HAGIS database, which contains pathological images. Our results show that, in addition to healthy retinal images, the proposed method performs well in extracting blood vessels from pathological images and is therefore comparable with state of the art methods.     

基于自适应融合的手掌静脉增强方法

针对掌脉轮廓不清晰，图像对比度低、亮度低，进而导致识别性能降低的现象，提出一种自适应融合的手掌静脉增强方法。首先，基于暗原色先验（DCP）去雾算法，根据掌脉图像变异系数自适应选择去雾系数，得到DCP增强图像，并且基于部分子块重叠直方图均衡（POSHE）算法得到POSHE增强图像；然后，将图像分为16个子块，依据图像灰度均值与标准差确定各子块权重；最后，根据各子块权重对DCP和POSHE增强图像进行自适应融合，得到最终增强图像。该方法既保留了DCP算法在增强图像对比度和亮度的同时不引入明显噪声的优点，又保留了POSHE算法在增强图像对比度和亮度的同时不损失局部细节的特点；同时，两者的自适应融合既解决了DCP图像阴影部分掌脉缺失现象，又削弱了POSHE产生的块效应。在对两个公开库和自建库分别进行的实验中，三个数据库的等错误率分别为0.0004、0.0472、0.0579，识别率分别为99.98%、94.27%、92.05%。实验结果表明，与现有的图像增强方法相比，该方法降低了等错误率，提高了识别精度。     

水下图像处理技术研究综述

随着自主式水下机器人的发展，水下探测技术成为新的研究热点。然而，吸收效应和散射效应导致水下获取的图像存在雾化和色彩偏差等缺陷。降质的水下图像在一定程度上降低了水下目标识别的准确性。为了改善水下图像质量，国内外学者对水下图像处理方法进行了深入研究。因水下图像处理方法对提升水下目标识别准确性具有良好的促进作用，故其具有重要的研究与分析价值。介绍了水下成像模型，分析了水下图像视觉质量下降的原理；根据水下物理成像模型将水下图像处理方法分为水下图像增强与水下图像复原，并分别对两类方法的研究现状进行分析与归纳；最后，总结与讨论了各类方法的优缺点，并展望了未来的发展方向。     

Underwater image enhancement with global–local networks and compressed-histogram equalization

Due to the light absorption and scattering, captured underwater images usually contain severe color distortion and contrast reduction. To address the above problems, we combine the merits of deep learning and conventional image enhancement technology to improve the underwater image quality. We first propose a two-branch network to compensate the global distorted color and local reduced contrast, respectively. Adopting this global–local network can greatly ease the learning problem, so that it can be handled by using a lightweight network architecture. To cope with the complex and changeable underwater environment, we then design a compressed-histogram equalization to complement the data-driven deep learning, in which the parameters are fixed after training. The proposed compression strategy is able to generate vivid results without introducing over-enhancement and extra computing burden. Experiments demonstrate that our method significantly outperforms several state-of-the-arts in both qualitative and quantitative qualities.     

Underwater image quality enhancement through integrated color model with Rayleigh distribution

The physical properties of water cause light-induced degradation of underwater images. Light rapidly loses intensity as it travels in water, depending on the color spectrum wavelength. Visible light is absorbed at the longest wavelength first. Red and blue are the most and least absorbed, respectively. Underwater images with low contrast are captured due to the degradation effects of light spectrum. Therefore, the valuable information from these images cannot be fully extracted for further processing. The current study proposes a new method to improve the contrast and reduce the noise of underwater images. The proposed method integrates the modification of image histogram into two main color models, Red–Green–Blue (RGB) and Hue-Saturation-Value (HSV). In the RGB color model, the histogram of the dominant color channel (i.e., blue channel) is stretched toward the lower level, with a maximum limit of 95%, whereas the inferior color channel (i.e., red channel) is stretched toward the upper level, with a minimum limit of 5%. The color channel between the dominant and inferior color channels (i.e., green channel) is stretched to both directions within the whole dynamic range. All stretching processes in the RGB color model are shaped to follow the Rayleigh distribution. The image is converted into the HSV color model, wherein the S and V components are modified within the limit of 1% from the minimum and maximum values. Qualitative analysis reveals that the proposed method significantly enhances the image contrast, reduces the blue-green effect, and minimizes under- and over-enhanced areas in the output image. For quantitative analysis, the test with 300 underwater images shows that the proposed method produces average mean square error (MSE) and peak signal to noise ratio (PSNR) of 76.76 and 31.13, respectively, which outperform six state-of-the-art methods.     

基于信号增强的缓慢泄漏检测方法

目前管道泄漏检测方法可有效检测突发泄漏，对于缓慢泄漏则存在检测灵敏度低、定位不准确等问题。基于此，提出了一种基于信号增强的缓慢泄漏检测方法。通过信号压缩（抽取及移位）克服缓慢泄漏压力信号下降平缓的缺点；根据声波信号具有波形尖锐突出、对突发泄漏敏感的优点，通过建立以压力为输入、虚拟声波为输出的声波信号变送器模型，将压力信号转换为声波信号，克服了泄漏压力信号容易被淹没在管道压力波动及背景噪声中的缺点，实现了缓慢泄漏信号的增强；利用临近插值方法重构虚拟声波信号，基于延时互相关分析实现了缓慢泄漏的准确定位。实验结果表明，该方法具有显著的信号增强效果和定位精度，实现了缓慢泄漏的准确检测。     

Robust image retrieval by cascading a deep quality assessment network

The performance of computer vision algorithms can severely degrade in the presence of a variety of distortions. While image enhancement algorithms have evolved to optimize image quality as measured according to human visual perception, their relevance in maximizing the success of computer vision algorithms operating on the enhanced image has been much less investigated. We consider the problem of image enhancement to combat Gaussian noise and low resolution with respect to the specific application of image retrieval from a dataset. We define the notion of image quality as determined by the success of image retrieval and design a deep convolutional neural network (CNN) to predict this quality. This network is then cascaded with a deep CNN designed for image denoising or super resolution, allowing for optimization of the enhancement CNN to maximize retrieval performance. This framework allows us to couple enhancement to the retrieval problem. We also consider the problem of adapting image features for robust retrieval performance in the presence of distortions. We show through experiments on distorted images of the Oxford and Paris buildings datasets that our algorithms yield improved mean average precision when compared to using enhancement methods that are oblivious to the task of image retrieval. ¹     

Flexible Piezoelectric Acoustic Sensors and Machine Learning for Speech Processing

Flexible piezoelectric acoustic sensors have been developed to generate multiple sound signals with high sensitivity, shifting the paradigm of future voice technologies. Speech recognition based on advanced acoustic sensors and optimized machine learning software will play an innovative interface for artificial intelligence (AI) services. Collaboration and novel approaches between both smart sensors and speech algorithms should be attempted to realize a hyperconnected society, which can offer personalized services such as biometric authentication, AI secretaries, and home appliances. Here, representative developments in speech recognition are reviewed in terms of flexible piezoelectric materials, self-powered sensors, machine learning algorithms, and speaker recognition.     

Diverse bio-sensing and therapeutic applications of plasmon enhanced nanostructures

Substantial advancements have been observed over the years in the research and development of Localized Surface Plasmon Resonance (LSPR). A variety of current and future applications involving anisotropic plasmonic nanoparticles include biosensors, photothermal therapies, photocatalysis, and various other fields. Amongst various other applications, plasmonic enhancements are deployed in Surface Enhanced Raman Spectroscopy (SERS) mediated bio-sensing, absorption spectroscopy based analyte quantification, and fluorescence spectroscopy-based biomolecular detection up to femtomolar level and even on the level of single molecules. LSPR based healthcare diagnostics and therapeutics have grown much faster than expected, with an increased number of published original research articles and reviews. Despite the extensive literature available, a comprehensive review with a focused emphasis on recent advances in the field of plasmonic particle anisotropy, plasmonic nanostructure, plasmonic coupling mediated enhanced LSPR intensity and their diverse applications in biosensing is needed. This article focuses on LSPR properties of anisotropic nanostructures like spherical gold nanoparticles (AuNP), gold nanorod (AuNR), gold nanostar (AuNs), gold nanorattles (AuNRT), gold nanoholes (AuNH), dimeric nanostructures and their role in plasmonic enhancements for targeted biosensing and therapeutic research. The contemporary state of the art biosensing development around SERS has also been discussed. A detailed literature analysis of recent development in micro-surgery, photothermal tumor killing, biosensor development for detection up to single molecule level, high-efficiency drug delivery are covered in this article. Furthermore, recent and advanced technologies including Spatially Offset Raman Spectroscopy (SORS), Surface Enhanced Resonance Raman Spectroscopy (SERRS), and Surface Enhanced Spatially Offset Raman Spectroscopy (SESORS) are presented citing their importance in biosensing. We complement this review article with relevant theoretical frameworks to understand finer nuances within the literature that is discussed.