VMD-DMD coupled data-driven approach for visual saliency in noisy images

Multimedia Tools and Applications - Human visual system is endowed with an innate capability of distinguishing the salient regions of an image. It do so even in the presence of noise and other...     

Two pyramid machine algorithms for edge detection in noisy binary images

This paper presents a pair of fast algorithms which detect edges in noisy binary images. The algorithms run on a parallel hierarchical cellular array processor called a pyramid machine. Example results are illustrated and the time required by each algorithm is discussed.     

An adaptive enhancement algorithm based on visual saliency for low illumination images

Applied Intelligence - In order to improve the brightness and contrast of low illumination color images and avoid over enhancement, an adaptive image enhancement algorithm based on visual saliency...     

Nonlinear restoration of noisy images

The restoration of images degraded by an additive white noise is performed by nonlinearly filtering a noisy image. The standard Wiener approach to this problem is modified to take into account the edge information of the image. Various filters of increasing complexity are derived. Experimental results are shown and compared to the standard Wiener filter results and other earlier attempts involving nonstationary filters.     

Bandwidth compression of noisy images

The effect of additive noise on Discrete Fourier Transform of pictorial data (2DFT) must be considered in several important problems such as filtering, enhancement and bandwidth compression. This paper investigates the statistical properties of the corrupted 2DFT coefficients, and the error involved in reconstruction when a subset of these coefficients is employed for the purpose. The latter being often the case, the present analysis provides a rational basis for frequency selection and filter specification. Application of an F-test quantifies the reliability of filtering or bandwidth compression.     

Estimation-detection of object boundaries in noisy images

Estimation of boundaries of objects in noisy images is considered when the objects and the background are statistically characterized. The noise is assumed white, additive, and Gaussian. Optimal recursive estimators in a joint estimation-detection context are derived. Applications to binary pictures are illustrated.     

Point reconstruction from noisy images

In this paper we treat the problem of determining optimally (in the least-squares sense) the 3D coordinates of a point, given its noisy images formed by any number of cameras of known geometry. The optimality criterion is determined by the covariance matrices associated with the images of the point. The covariance matrices are not restricted to be positive definite but are allowed to be singular. Thus, image points constrained to lie along straight lines can be handled as well. Estimation of the covariance of the reconstructed point is provided.The often appearing two-camera stereo case is treated in detail. It is shown in this case that, under reasonable conditions, the main step of the reconstruction reduces to finding the unique zero of a sixth degree polynomial in the interval (0, 1).The authors are listed in random order.     

Backlight dimming based on saliency map acquired by visual attention analysis

Displays have been used in various applications from mobile phones to tablets, and the low power consumption is one of their most important factors. Backlight dimming is the most promising technique to achieve this because it significantly reduces the display power by controlling only the transmittance of liquid crystal. This paper proposes a new backlight dimming algorithm using visual attention analysis. Conventional algorithms have a serious saturation error in some images when performing backlight dimming, thereby degrading image quality. In contrast, the proposed algorithm analyzes image characteristics based on the saliency map, which considers human visual attention. It then truncates the meaningless information of the saliency map using an adaptive saliency level selection approach and calculates the maximum amount of saturation error that humans will not perceive. In addition, the proposed algorithm defines the objective function and computes the optimal starting gray level in that function to calculate the saturation error. Simulation results show that the proposed algorithm using the adaptive saliency level selection approach performs best. In addition, the average peak signal-to-noise ratio of the proposed algorithm was up to 3.762 dB higher than that of the conventional algorithm while slightly increasing the power consumption.     

利用视觉显著性的图像分割方法

提出一种利用视觉显著性对图像进行分割的方法。首先提取图像的底层视觉特征,从局部显著性、全局显著性和稀少性3个方面计算各特征图像中各像素的视觉显著性,得到各特征显著图;对各特征显著图进行综合,生成最终的综合显著图。然后对综合显著图进行阈值分割,得到二值图像,将二值图像与原始图像叠加,将前景和背景分离,得到图像分割结果。在多幅自然图像上进行实验验证,并给出相应的实验结果和分析。实验结果表明,该方法正确有效,具有和人类视觉特性相符合的分割效果。     

Neural-swarm visual saliency for path following

This paper extends an existing saliency-based model for path detection and tracking so that the appearance of the path being followed can be learned and used to bias the saliency computation process. The goal is to reduce ambiguities in the presence of strong distractors. In both original and extended path detectors, neural and swarm models are layered in order to attain a hybrid solution. With generalisation to other tasks in mind, these detectors are presented as instances of a generic neural-swarm layered architecture for visual saliency computation. The architecture considers a swarm-based substrate for the extraction of high-level perceptual representations, given the low-level perceptual representations extracted by a neural-based substrate. The goal of this division of labour is to ensure parallelism across the vision system while maintaining scalability and tractability. The proposed model is shown to exhibit, at 20 Hz, a 98.67% success rate on a diverse data-set composed of 39 videos encompassing a total of 29,789 640 × 480 frames. An open source implementation of the model, fully encapsulated as a node of the Robotics Operating System (ROS), is available for download.     

多对象图像数据集建立及显著性检测算法评估

图像视觉显著性检测算法在已有数据集上已经取得很好的结果,但是目前的多个数据集存在两个严重的问题:首先,数据集中的图像以只包含一个显著对象的图像为主;其次,在建立显著对象标注结果的过程中,忽略了用户对同一幅图像中包含的多个显著对象的不同认知。上述问题导致了在已有数据集上对显著性检测算法进行评估,不能体现算法在实际应用中的真实效果。为此,提出体现用户认知的多显著对象图像标注方法,首先设计并实现辅助软件,收集用户对各显著对象的重要程度的认知情况,包括显著区域与相应的重要程度;然后融合收集的多用户数据,绘制出以灰度图为表现形式的显著对象标注结果,并通过灰度值体现多用户对于每个显著对象的认知情况。基于改进的显著对象标注方法,建立了一个包含1000幅多显著对象图像的数据集,并为每幅图像提供了体现用户认知的显著对象标注结果。对10种具有代表性的显著性检测算法在已有数据集和建立的数据集上的性能进行了比较。实验结果表明,这些显著性检测算法在建立的数据集上的性能有大幅度的降低,例如受试者工作特征曲线下面积(ROC-AUC)评估参数的最大降幅超过了0.5,这证实了已有数据集存在的问题及建立新数据集的需求,同时指出显著性检测算法在处理包含多显著对象的复杂图像上存在的不足。     

Towards reliable object detection in noisy images

Image noise is a common problem frequently caused by insufficient lighting, low-quality cameras, image compression and other factors. While low image quality is expected to degrade results of visual recognition, most of the current methods and benchmarks for object recognition, such as Pascal Visual Object Classes Challenge and Microsoft Common Objects in Context Challenge, focus on relatively high-quality images. Meanwhile, object recognition in noisy images is a common problem in surveillance and other domains. In this work we address object detection in noisy images and propose a novel low-cost method for image denoising. When combined with the standard Deformable Parts Model and Regions with Convolutional Neural Network object detectors, our method shows improvements of object detection under varying levels of image noise. We present a comprehensive experimental evaluation and compare our method to other denoising techniques as well as to standard detectors re-trained on noisy images. Results are presented for the common Pascal Visual Object Classes benchmark for object detection and KAIST Multispectral Pedestrian Detection Benchmark with the real noise presence in night images.     

Infrared and visible images fusion using visual saliency and optimized spiking cortical model in non-subsampled shearlet transform domain

Multimedia Tools and Applications - Aiming at some problems in existing infrared and visible image fusion methods such as edge blurring, low contrast, loss of details, a novel fusion scheme based...     

A novel remote-sensing image fusion method based on hybrid visual saliency analysis

ABSTRACT

In practical applications, different regions in images may practically have different demands for the spatial and spectral resolution. However, most existing methods execute a unified fusion processing of the whole image with no consideration of such diverse demands. In this article, a new fusion method for remote-sensing images based on saliency analysis is proposed for addressing the issue. By introducing the hybrid visual saliency analysis, regions in the panchromatic (Pan) image and multispectral image would be automatically partitioned into two parts: salient and non-salient regions. Then, a subregion fusion strategy is applied to fuse the non-salient and salient regions, respectively. As for salient regions such as residential areas and roads, the adaptive intensity–hue–saturation (adaptive IHS) method is implemented for its merit of effective improvement in spatial quality. For non-salient regions such as farmland, forest, and grassland, the dual-tree complex wavelet transform is used in the process of spatial detail extraction, and the combination coefficients yielded by the adaptive IHS method are integrated to suppress the spectral distortion. Experimental results demonstrate that our proposal provides state-of-the-art performance as well as achieves a better balance between spatial injection and spectral maintenance in different regions.     

Machine learning-based framework for saliency detection in distorted images

Multimedia Tools and Applications - Visual saliency detection is useful in carrying out image compression, image segmentation, image retrieval, and other image processing applications. Majority of...     

On visual BMI analysis from facial images

Automatically assessing body mass index (BMI) from facial images is an interesting and challenging problem in computer vision. Facial feature extraction is an important step for visual BMI estimation. This work studies the visual BMI estimation problem based on the characteristics and performance of different facial representations, which has not been well studied yet. Various facial representations, including geometry based representations and deep learning based, are comprehensively evaluated and analyzed from three perspectives: the overall performance on visual BMI prediction, the redundancy in facial representations and the sensitivity to head pose changes. The experiments are conducted on two databases: a new dataset we collected, called the FIW-BMI and an existing large dataset Morph II. Our studies provide some deep insights into the facial representations for visual BMI analysis: 1) The deep model based methods perform better than geometry based methods. Among them, the VGG-Face and Arcface show more robustness than others in most cases; 2) Removing the redundancy in VGG-Face representation can increase the accuracy and efficiency in BMI estimation; 3) Large head poses lead to low performance for BMI estimation. The Arcface, VGG-Face and PIGF are more robust than the others to head pose variations.     

Bayesian feature evaluation for visual saliency estimation

This paper presents a computational method of feature evaluation for modeling saliency in visual scenes. This is highly relevant in visual search studies since visual saliency is at the basis of visual attention deployment. Visual saliency can also become important in computer vision applications as it can be used to reduce the computational requirements by permitting processing only in those regions of the scenes containing relevant information. The method is based on Bayesian theory to describe the interaction between top-down and bottom-up information. Unlike other approaches, it evaluates and selects visual features before saliency estimation. This can reduce the complexity and, potentially, improve the accuracy of the saliency computation. To this end, we present an algorithm for feature evaluation and selection. A two-color conjunction search experiment has been applied to illustrate the theoretical framework of the proposed model. The practical value of the method is demonstrated with video segmentation of instruments in a laparoscopic cholecystectomy operation.     

A novel change detection approach based on visual saliency and random forest from multi-temporal high-resolution remote-sensing images

ABSTRACT

This article presents a novel change detection (CD) approach for high-resolution remote-sensing images, which incorporates visual saliency and random forest (RF). First, highly homogeneous and compact image super-pixels are generated using super-pixel segmentation, and the optimal segmentation result is obtained through image superimposition and principal component analysis. Second, saliency detection is used to guide the search of interest regions in the initial difference image obtained via the improved robust change vector analysis algorithm. The salient regions within the difference image that correspond to the binarized saliency map are extracted, and the regions are subject to the fuzzy c-means (FCM) clustering to obtain the pixel-level pre-classification result, which can be used as a prerequisite for super-pixel-based analysis. Third, on the basis of the optimal segmentation and pixel-level pre-classification results, different super-pixel change possibilities are calculated. Furthermore, the changed and unchanged super-pixels that serve as the training samples are automatically selected. The spectral features and Gabor features of each super-pixel are extracted. Finally, super-pixel-based CD is implemented by applying RF based on these samples. Experimental results on Quickbird, Ziyuan 3 (ZY3), and Gaofen 2 (GF2) multi-spectral images show that the proposed method outperforms the compared methods in the accuracy of CD, and also confirm the feasibility and effectiveness of the proposed approach.     

Linear granulometric moments of noisy binary images

The first and second moments of the granulometric pattern spectrum are expressed for a random binary image that is formed either as the union of a deterministic signal with random point noise or as the set subtraction of random point noise from a deterministic signal. The granulometry is generated by a vertical (or horizontal) linear structuring element, and there are no constraints placed on the structure of the uncorrupted signal. Because the noise is random, the image on which the granulometry is run is random. Hence the pattern spectrum is a random function with random-variable moments. For both the union and subtractive cases, expressions are found for the expectation of the pattern-spectrum mean and variance, where the expectation is relative to the noise intensity. In each case a recursive formula is obtained for the key expression.