Gaussian Image Research Articles

Pre-whitened matched filter and convolutional neural network based model observer performance for mass lesion detection in non-contrast breast CT.

Mathematical model observers have been shown to reasonably predict human observer performance and are useful when human observer studies are infeasible. Recently, convolutional neural networks (CNNs) have also been used as substitutes for human observers, and studies have shown their utility as an optimal observer. In this study, a CNN model observer is compared to the pre-whitened matched filter (PWMF) model observer in detecting simulated mass lesions inserted into 253 acquired breast computed tomography (bCT) images from patients imaged at our institution. To compare CNN and PWMF model observers for detecting signal-known-exactly (SKE) location-known-exactly (LKE) simulated lesions in bCT images with real anatomical backgrounds, and to use these model observers collectively to optimize parameters and understand trends in performance with breast CT. Spherical lesions with different diameters (1, 3, 5, 9mm) were mathematically inserted into reconstructed patient bCT image data sets to mimic 3D mass lesions in the breast. 2D images were generated by extracting the center slice along the axial dimension or by slice averaging across adjacent slices to model thicker sections (0.4, 1.2, 2.0, 6.0, 12.4, 20.4mm). The role of breast density was retrospectively studied using the range of breast densities intrinsic to the patient bCT data sets. In addition, mass lesions were mathematically inserted into Gaussian images matched to the mean and noise power spectrum of the bCT images to better understand the performance of the CNN in the context of a known ideal observer (the PWMF). The simulated Gaussian and bCT images were divided into training and testing data sets. Each training data set consisted of 91 600 images, and each testing data set consisted of 96 000 images. A CNN and PWMF was trained on the Gaussian training images, and a different CNN and PWMF was trained on the bCT training images. The trained model observers were tested, and receiver operating characteristic (ROC) curve analysis was used to evaluate detection performance. The area under the ROC curve (AUC) was the primary performance metric used to compare the model observers. In the Gaussian background, the CNN performed essentially identically to the PWMF across lesion sizes and section thicknesses. In the bCT background, the CNN outperformed the PWMF across lesion size, breast density, and most section thicknesses. These findings suggest that there are higher-order features in bCT images that are harnessed by the CNN observer but are inaccessible to the PWMF. The CNN performed equivalently to the ideal observer in Gaussian textures. In bCT background, the CNN captures more diagnostic information than the PWMF and may be a more pertinent observer when conducting optimal performance studies in breast CT images.

Sub-minute acquisition with deep learning-based image filter in the diagnosis of colorectal cancers using total-body 18F-FDG PET/CT

BackgroundThis study aimed to retrospectively evaluate the feasibility of total-body 18F-FDG PET/CT ultrafast acquisition combined with a deep learning (DL) image filter in the diagnosis of colorectal cancers (CRCs).MethodsThe clinical and preoperative imaging data of patients with CRCs were collected. All patients underwent a 300-s list-mode total-body 18F-FDG PET/CT scan. The dataset was divided into groups with acquisition durations of 10, 20, 30, 60, and 120 s. PET images were reconstructed using ordered subset expectation maximisation, and post-processing filters, including a Gaussian smoothing filter with 3 mm full width at half maximum (3 mm FWHM) and a DL image filter. The effects of the Gaussian and DL image filters on image quality, detection rate, and uptake value of primary and liver metastases of CRCs at different acquisition durations were compared using a 5-point Likert scale and semi-quantitative analysis, with the 300-s image with a Gaussian filter as the standard.ResultsAll 34 recruited patients with CRCs had single colorectal lesions, and the diagnosis was verified pathologically. Of the total patients, 11 had liver metastases, and 113 liver metastases were detected. The 10-s dataset could not be evaluated due to high noise, regardless of whether it was filtered by Gaussian or DL image filters. The signal-to-noise ratio (SNR) of the liver and mediastinal blood pool in the images acquired for 10, 20, 30, and 60 s with a Gaussian filter was lower than that of the 300-s images (P < 0.01). The DL filter significantly improved the SNR and visual image quality score compared to the Gaussian filter (P < 0.01). There was no statistical difference in the SNR of the liver and mediastinal blood pool, SUVmax and TBR of CRCs and liver metastases, and the number of detectable liver metastases between the 20- and 30-s DL image filter and 300-s images with the Gaussian filter (P > 0.05).ConclusionsThe DL filter can significantly improve the image quality of total-body 18F-FDG PET/CT ultrafast acquisition. Deep learning-based image filtering methods can significantly reduce the noise of ultrafast acquisition, making them suitable for clinical diagnosis possible.

Devising a method for determining the aperture brightness of an object using a typical shape of its image

The object of this study is the aperture brightness of the image of the object, which has a variety of typical shapes in the frames of the series. It directly depends on the stability of the shooting conditions of the objects under study. Thus, determining the exact aperture brightness of an object in the frame becomes more difficult. For this purpose, a method was devised for determining the aperture brightness of an object using the typical shape of its image on a series of frames. This method is based on the formation of a typical shape of a digital image of an object based on data from all frames of the series. The typical shape makes it possible to take into account the peculiarities of the formation of the image of the object on each frame of the series. Based on this, a more accurate estimate of the initial approximation of the parameters of all Gaussian images of the object is performed. In addition, the adaptation of the method to the standard shape makes it possible to perform a more accurate assessment of the aperture brightness of the object in comparison with the analytically defined profile. An estimate of the aperture brightness of the object was derived using the least squares method. Due to minimization using the Levenberg-Marquardt algorithm, the use of the method improved identification with reference objects and reduced the number of false positives. The study showed a decrease in the standard deviation of frame identification errors by 5–7 times when using a typical shape of a digital image. The devised method for determining an object's aperture brightness using its image's typical shape was tested in practice within the framework of the CoLiTec project. It was implemented in the intra-frame processing unit of the CoLiTecVS software for the automated construction of brilliance curves of the studied variable stars. Owing to the use of the CoLiTecVS software and the proposed computational method implemented in it, more than 700,000 measurements of various objects under study were successfully processed and identified

Deep learning-based denoising in projection-domain and reconstruction-domain for low-dose myocardial perfusion SPECT

BackgroundLow-dose (LD) myocardial perfusion (MP) SPECT suffers from high noise level, leading to compromised diagnostic accuracy. Here we investigated the denoising performance for MP-SPECT using a conditional generative adversarial network (cGAN) in projection-domain (cGAN-prj) and reconstruction-domain (cGAN-recon). MethodsSixty-four noisy SPECT projections were simulated for a population of 100 XCAT phantoms with different anatomical variations and 99mTc-sestamibi distributions. Series of LD projections were obtained by scaling the full dose (FD) count rate to be 1/20 to 1/2 of the original. Twenty patients with 99mTc-sestamibi stress SPECT/CT scans were retrospectively analyzed. For each patient, LD SPECT images (7/10 to 1/10 of FD) were generated from the FD list mode data. All projections were reconstructed by the quantitative OS-EM method. A 3D cGAN was implemented to predict FD images from their corresponding LD images in the projection- and reconstruction-domain. The denoised projections were reconstructed for analysis in various quantitative indices along with cGAN-recon, Gaussian, and Butterworth-filtered images. ResultscGAN denoising improves image quality as compared to LD and conventional post-reconstruction filtering. cGAN-prj can further reduce the dose level as compared to cGAN-recon without compromising the image quality. ConclusionsDenoising based on cGAN-prj is superior to cGAN-recon for MP-SPECT.

Lattice Boltzmann model for diffusion equation with reduced truncation errors: Applications to Gaussian filtering and image processing

Several image processing approaches require blurring an image using the Gaussian filter. In the present study the filtering method based on a novel four-velocity lattice Boltzmann model with non-local collision kernel is proposed. By applying the Chapman–Enskog expansion up to super-Burnett terms it is shown that the present scheme is equivalent to the diffusion equation with relatively small hyper-viscous defects (truncation errors). Compared to the conventional lattice Boltzmann model the proposed scheme has a small computational overhead but significantly improves accuracy due to the reduced hyper-viscosity. To validate the model the following test problems are considered: diffusion in time of a Gaussian bell, sinusoidal wave, image filtering, feature detection with the application of the modified SIFT method. The numerical experiments indicate that the novel model gives more precise results than the standard lattice Boltzmann model for non-small values of diffusion coefficient and in several cases its computational cost is competitive with OpenCV blurring techniques. In addition, the model can be efficiently used in feature detection algorithms like SIFT.

Deep expectation-maximization network for unsupervised image segmentation and clustering

Unsupervised learning, such as unsupervised image segmentation and clustering, are fundamental tasks in image representation learning. In this paper, we design a deep expectation-maximization (DEM) network for unsupervised image segmentation and clustering. It is based on the statistical modeling of image in its latent feature space by Gaussian mixture model (GMM), implemented in a novel deep learning framework. Specifically, in the unsupervised setting, we design an auto-encoder network and an EM module over the image latent features, for jointly learning the image latent features and GMM model of the latent features in a single framework. To construct the EM-module, we unfold the iterative operations of EM algorithm and the online EM algorithm in fixed steps to be differentiable network blocks, plugged into the network to estimate the GMM parameters of the image latent features. The proposed network parameters can be end-to-end optimized using losses based on log-likelihood of GMM, entropy of Gaussian component assignment probabilities and image reconstruction error. Extensive experiments confirm that our proposed networks achieve favorable results compared with several state-of-the-art methods in unsupervised image segmentation and clustering.

Hierarchical reconstruction of 3D well-connected porous media from 2D exemplars using statistics-informed neural network

The relationships between porous microstructures and transport properties are of fundamental importance in various scientific and engineering applications. Due to the intricacy, stochasticity and heterogeneity of porous media, reliable characterization and modeling of transport properties often require a complete dataset of internal microstructure samples. However, it is often an unbearable cost to acquire sufficient 3D digital microstructures by purely using microscopic imaging systems. This paper presents a machine learning-based technique to hierarchically reconstruct 3D well-connected porous microstructures from one isotropic or several anisotropic low-cost 2D exemplar(s). To compactly characterize the large-scale microstructural features, a Gaussian image pyramid is built for each 2D exemplar. Local morphology patterns are collected from the Gaussian image pyramids, and then they serve as the training data to embed the 2D morphological statistics into feed-forward neural networks at multiple length levels. By using a specially-developed morphology integration scheme, the 3D morphological statistics at different levels can be inferred from the statistics-informed neural networks. Gibbs sampling is adopted to hierarchically reconstruct 3D microstructures by using multi-level 3D morphological statistics, where the large-scale, regional and local morphological patterns are statistically generated and successively added to the same 3D random field. The proposed method is tested on a series of porous media with distinct morphologies, and the statistical equivalence between the reconstructed and the real microstructures is systematically evaluated by comparing morphological descriptors and transport properties. The results demonstrate that the proposed 2D-to-3D microstructure reconstruction method is a universal and efficient approach to generating morphologically and physically realistic samples of porous media.

Inter-seasonal characterization and correlation of Surface Urban Heat Island (SUHI) and Canopy Urban Heat Island (CUHI) in the urbanized environment of Delhi

The Intergovernmental Panel on Climate Change's (IPCC) 6th assessment report warns India about excessive heat. As India's population, urbanization, built-up regions, and concrete expand, vulnerable people may face higher heat risks. Delhi, India's capital and second-largest city, is noted for its excessive heat and air pollution. Heat in Delhi is spatially distributed and varies in magnitude due to built-up area density, the absence of a natural landscape, and other geographical aspects. To pinpoint the hotspot region, the study attempts to characterize Surface Urban Heat Islands (SUHIs) and Canopy Urban Heat Islands (CUHIs). Temporal and seasonal assessment of LST (Land Surface Temperature) using MODIS (Moderate Resolution Imaging Spectroradiometers) data and Ambient Temperature (AT) data from 39 Central Pollution Control Board (CPCB) monitoring stations with a representative radius of 1–5 km used to explore the inter-seasonal variation of SUHI and CUHI over the city in terms of magnitude and extent for 2019, 2020, and 2021. To describe SUHI and CUHI, a two-dimensional Gaussian fitting-based image analysis method is developed. The results show that the region's greatest SUHI magnitude is 2.36 °C in the monsoon season and 2.09 °C in the spring season. In comparison, the maximum CUHI magnitude is 4.74 °C and 3.15 °C in the Autumn and Winter seasons, respectively. The impact of SUHI can be experienced up to 82 km2 from the critical location in Winter and CUHI has 67.83 km2 in Summer. Inter seasonal comparison matrix between SUHI and CUHI demonstrates a positive but modest association due to considerable seasonal parameter fluctuation. UHI hotspot zone is at the centre of the urban with little seasonal fluctuation. South-West (SW) is a hotspot for SUHI and CUHI during the monsoon and spring seasons due to its dense residential and industrial areas. During winter, SUHI indicates SW as the critical zone and CUHI depicts NW. The finding shows that Delhi has a longer duration of hot weather. This study could help to understand UHI in Delhi at the local scale. The study shows UHI's seasonality and potential drivers. Critical hotspots and seasonal shifts should be used in urban development to improve Delhi's thermal environment.

Mobile visual media design based on artificial intelligence and augmented reality technology

Abstract In recent years, with the development of computer and network technologies, digital video and image information is emerging more and more, the information era based on multimedia information services is coming to us, and people’s demand for visual media content such as video and images is becoming more and more extensive. In this paper, we use the Gaussian mixture model pixel detection technology method and image detection technology to study visual media to meet the various needs of users. SVM is used to perform classification tests on 3688 Chinese painting images and 4827 general images. The result is that the positive detection rate of national painting images is 87.25%, and the rate of non-national painting images being wrongly detected as national painting images is 7.63%. This indicates that the CCV feature also has a relatively good discriminative property for national painting images. The focus is on providing technical methods for the design of various visual media shapes such as sports and art, thus providing solutions to help users better access and use the digital multimedia information they are interested in.

Unburnt carbon estimation through flame image and gauss process regression

ABSTRACT The presence of unburned carbon in coal-burning systems undoubtedly causes a loss in the amount of energy that can be obtained from the system, and also reveals an inadequacy in terms of the usability of the ashes. The expensiveness of existing unburned carbon prediction methods is one of the reasons why these technologies cannot be used. This situation requires working on alternative non-combustible carbon technologies. In this paper, a new approach is presented for estimating unburned carbon in a small-scale coal burner system using the Gaussian regression model and CCD camera-acquired flame image. The proposed approach evaluates brightness, fluctuation amplitude, area, and radiation signal properties of the flame image. The proposed non-combustible carbon estimation technique does not require prior knowledge of CCD camera features. In the feature acquisition phase, results were obtained for each natural component of the flame image in RGB colour space separately, in pairs, all together and for three artificial colour channels (grey image). With the proposed method, the unburned carbon estimation was obtained with an accuracy of R = 0.9664 when all colour channels of the RGB image were used together. This result shows that unburned carbon can be estimated from the instantaneous flame images obtained with the CCD camera.

Multiscale structural feature transform for multi-modal image matching

Robust and reliable multi-modal image matching is an essential and challenging technique in the multi-modality involved scenarios for image fusion, image mosaic, and visual navigation. Existing image matching methods cannot simultaneously solve the scale and rotation distortion caused by different viewpoints and the nonlinear radiation distortion caused by different imaging environments or mechanisms. To address these problems, a multiscale structural feature transform method is proposed for multi-modal image matching, which consists of three blocks: multiscale structural feature detector, symmetric log-polar descriptor, and point matching with position correction. The multiscale structural feature detector constructs the phase congruency map based Difference of Gaussian image pyramid for the scale-invariant feature points detection to tackle the nonlinear radiation distortion. The feature points are detected within 10-neighborhood instead of 26-neighborhood to produce sufficient points on the phase congruency map with very sparse texture. The symmetric log-polar descriptor utilizes the multiscale structure principal direction and symmetric bins to improve the robustness of the descriptor to rotation and nonlinear radiation distortion. At last, the point matching with position correction corrects the position of feature points to compensate the position offset between different images caused by the scale and rotation distortion and the nonlinear radiation distortion, which increases the number of correct matches and improves the transformation consistency between matched points. In the experiment, the proposed method in this paper is compared with seven state-of-the-art methods. Experimental results verify the superiority of the proposed method and its robustness to the scale and rotation distortion and the nonlinear radiation distortion. In addition, we demonstrate the effectiveness of the proposed three blocks by the ablation experiments.

Feasibility of Dedicated Breast Positron Emission Tomography Image Denoising Using a Residual Neural Network.

This study aimed to create a deep learning (DL)-based denoising model using a residual neural network (Res-Net) trained to reduce noise in ring-type dedicated breast positron emission tomography (dbPET) images acquired in about half the emission time, and to evaluate the feasibility and the effectiveness of the model in terms of its noise reduction performance and preservation of quantitative values compared to conventional post-image filtering techniques. Low-count (LC) and full-count (FC) PET images with acquisition durations of 3 and 7 minutes, respectively, were reconstructed. A Res-Net was trained to create a noise reduction model using fifteen patients' data. The inputs to the network were LC images and its outputs were denoised PET (LC + DL) images, which should resemble FC images. To evaluate the LC + DL images, Gaussian and non-local mean (NLM) filters were applied to the LC images (LC + Gaussian and LC + NLM, respectively). To create reference images, a Gaussian filter was applied to the FC images (FC + Gaussian). The usefulness of our denoising model was objectively and visually evaluated using test data set of thirteen patients. The coefficient of variation (CV) of background fibroglandular tissue or fat tissue were measured to evaluate the performance of the noise reduction. The SUVmax and SUVpeak of lesions were also measured. The agreement of the SUV measurements was evaluated by Bland-Altman plots. The CV of background fibroglandular tissue in the LC + DL images was significantly lower (9.102.76) than the CVs in the LC (13.60 3.66) and LC + Gaussian images (11.51 3.56). No significant difference was observed in both SUVmax and SUVpeak of lesions between LC + DL and reference images. For the visual assessment, the smoothness rating for the LC + DL images was significantly better than that for the other images except for the reference images. Our model reduced the noise in dbPET images acquired in about half the emission time while preserving quantitative values of lesions. This study demonstrates that machine learning is feasible and potentially performs better than conventional post-image filtering in dbPET denoising.

Super-resolution Raman imaging towards visualisation of nanoplastics.

Confocal Raman imaging can potentially identify and visualise microplastics and nanoplastics, but the imaging lateral resolution is hampered by the diffraction of the laser, making it difficult to analyse nanoplastics that are smaller than the laser spot and the lateral resolution limit (λ/2NA). Fortunately, once a nanoplastic is scanned to collect the spectrum via a position/pixel array as a spectrum matrix, akin to a hyperspectral matrix, the nanoplastic can be imaged by mapping the spectrum intensity as a pattern that is transcended axially and can be fitted as a 2D Gaussian surface. The Gaussian fitting and image re-construction by deconvolution can precisely predict the nanoplastic's position and approximate size, and potentially enhance the signal intensity. Several algorithms are also employed to decode the spectrum matrix, to improve the Raman images before the subsequent image re-construction. Overall, general confocal microscopy can also break through the diffraction limit of the excitation light by using algorithms, resulting in super-resolution Raman imaging to capture nanoplastics.

Edge preserving image denoising with deep networks and ensemble convolutional neural network architectures

In the field of visual perception, the edges of images tend to be rich in effective visual stimuli, which contribute to the neural network's understanding of various scenes. Image smoothing is an image processing method used to highlight the wide area, low-frequency components, and main part of the image or to suppress image noise and high-frequency interference components, which can make the image's brightness smooth and gradual, reduce the abrupt gradient, and improve the image quality. Reducing noise is treated as one of the important problems in image processing. At the same time, preserving the edges of objects is of critical importance to protect the visual appearance of the objects. Deep networks have marked a trend in computer vision applications and this paper presents a customized Gaussian noise minimizing network with edge preserving filers. Here, ensemble architecture of convolutional neural network is used in minimizing the Gaussian noise and image denoising. The ensemble architecture is combined with VGG-19 and Xception design of CNN. The ensemble Convolutional Neural Networks (CNNs) are classified for Gaussian noisy images, real noisy images, blind denoising, and hybrid noisy images, representing the combination of noisy, blurred, and lowresolution images. Following the classification, motivations and principles of various deep learning methods are analyzed. Subsequently, a comparison of state-of-the-art methods on public denoising datasets is conducted, considering both quantitative and qualitative analyses. The experimental analysis is carried out in terms of PSNR, accuracy, precision, recall and F-measure.

Automated hair removal in dermoscopy images using shallow and deep learning neural architectures

Removing hair from digital dermoscopy images is occasionally a necessary step before further analysis is applied to the images. This work considers two machine learning approaches that segment the hair pixels from dermoscopy images. Subsequently, morphological post-processing is applied to refine the segmented hair and an image inpainting algorithm replaces the hair pixels with values based on the surrounding image structures. The first hair segmentation approach combines pixel-wise features extracted using the well-known Gaussian image pyramid with a traditional shallow multilayer perceptron (MLP-ANN), to detect hair pixels in images. The second approach uses a deep neural convolutional Encoder – Decoder (ED) network to segment hair. Both hair segmentation methods (MLP-ANN and ED) are trained with a set of 32 dermoscopy images with manually annotated hair, whereas the MLP-ANN dataset is constructed in a pixel-wise manner.Both proposed methods underwent three different assessments. First a set of 50 images with a-priori known hair is used for hair segmentation evaluation. Secondly, a set of 13 different dermoscopy images with hair added using a suitably trained Generative Adversarial Network -GAN- are used to assess the quality of hair removal that generates the hair-free image, in terms of several error metrics with respect to the original hair-free image. Finally, both proposed hair segmentation methods (MLP-ANN and ED) are applied on a set of 200 hair and hair-free images, which is used for training an image classifier to recognize melanoma against nevi lesions and the improvement in the image classification accuracy is measured. Comparative results against several other state-of-the-art hair removal techniques are also presented.Results show that in terms of hair removal, both the proposed hair removal techniques outperform the best performing of the state-of-the-art methods under comparison, in terms of several error metrics. Considering the effect of hair removal on melanoma image classification, the application of both MLP-ANN and ED increases the accuracy of melanoma classification. In all assessments, the ED was consistently the best performer. The statistical significance of the findings is also established.

Spatial and temporal information fusion for human action recognition via Center Boundary Balancing Multimodal Classifier

This paper proposes a novel multimodal data classifier named Center Boundary Balancing Multimodal Classifier (CBBMC) to fuse and classify the spatial and temporal descriptors for recognizing human actions from depth video sequences. CBBMC is a composite algorithm integrating feature fusion and feature classification, in which Center Boundary Balancing Projection (CBBP) is used to balance the center and boundary information of feature class spaces. In order to solve the problem of multimodal information redundancy and isolation, two feature selection and fusion schemes of CBBMC based on embedded feature selection are presented. Moreover, two new action descriptors called Gaussian Pyramid Depth Motion Images (GP-DMI) and Depth Temporal Maps (DTM) are introduced to capture the multi-scale spatial and fine-grained temporal information of human activities. Finally, we present an effective spatial and temporal information fusion framework based on CBBMC for human action recognition. In order to evaluate the performance of the proposed approach, extensive experiments are conducted. The proposed method achieved impressive results on four benchmark datasets, namely MSR Action3D (96.33%), UTD-MHAD (94.41%), DHA (95.65%), and NTU RGB+D (83.31% cross-subject and 87.66% cross-view), even though only the depth modality was used. The experimental results demonstrate the effectiveness of our method.

Improving longitudinal resolution of Airy beams two-photon volume imaging with fluorescence lifetime imaging

The non-diffracting Airy beams have been reported to increase acquisition speed by projecting the axially significantly elongated information to a two-dimensional (2D) image. Due to the elongated focal length, an Airy volumetric image can cover information as much as stack images with multiple Gaussian images at different imaging depths, which significantly increases the volumetric imaging rate. However, the projection of three-dimensional (3D) structures on a 2D plane is usually indistinguishable, especially those that overlap along the axis. Therefore, we proposed an axially resolved volume imaging technique based on a two-photon fluorescence lifetime microscopy imaging (FLIM) by using Airy beams.

A Correction Algorithm of the Sampling Effect and Its Application in High-Precision Astrometry

The sampling effect of the imaging acquisition device is an integration of the input signal within the pixel, resulting in an additional error in the pixel value. Additionally, a sampler with asymmetric intra-pixel quantum efficiency leads to position errors in high-precision astrometry. This paper proposes a model for the integral sampling process. An algorithm that solves the sampling effect, as well as the position error with high accuracy, is also provided. This algorithm provides an accuracy increase of 106 for Gaussian images with a uniform integral sampler. The accuracy limit of the Gaussian image comes from the truncation error. Also, this algorithm provides about 4 times accuracy improvement by eliminating the systematic error caused by the integral sampler with asymmetric intra-pixel quantum efficiency.

Single underwater image haze removal with a learning-based approach to blurriness estimation

Underwater images are usually degraded due to light scattering and absorption. To recover the scene radiance of degraded underwater images, a new haze removal method is presented by incorporating a learning-based approach to blurriness estimation with the image formation model. Firstly, the image blurriness is estimated with a linear model trained on a set of selected grayscale images, the average Gaussian images and blurriness images. With the estimated image blurriness, three intermediate background lights (BLs) are computed to obtain the synthesized BL. Then the scene depth is calculated by using the estimated image blurriness and BL to construct a transmission map and restore the scene radiance. Compared with other haze removal methods, haze in degraded underwater images can be removed more accurately with our proposed method. Moreover, visual inspection, quantitative evaluation and application test demonstrate that our method is superior to the compared methods and beneficial to high-level vision tasks.

SPAD monitoring of saline vegetation based on Gaussian mixture model and UAV hyperspectral image feature classification

The chlorophyll content of saline vegetation can indirectly reflect salinization. Rapid and non-destructive capture of chlorophyll content of saline vegetation at a regional scale is essential for saline soil improvement and sustainability of saline agriculture. However, traditional soil–plant analyzer development (SPAD) monitoring based on SPAD-502 is carried out at the leaf scale, which does not allow rapid access to SPAD information for the whole region. In this study, we proposed that post-hyperspectral classification based on spectral differences that could contribute to enhanced estimation of SPAD in saline vegetation. To test this proposal, we partitioned the hyperspectral images using a Gaussian mixture model. Then, estimation models based on spectral partition and full-sample SPAD were developed using in situ observation data and a random forest model, respectively. Finally, the unmanned aerial vehicles (UAV) hyperspectral images were used as the input data source to digitally map the SPAD of saline vegetation in the region, using the prediction model. The results indicated that there were significant intensity and shape differences in the spectral reflectance characteristics of saline vegetation under different clusters. The SPAD prediction model, based on spectral feature partition, performed significantly better than the full sample. The SPAD maps of saline vegetation before and after clustering displayed similar spatial distribution models, but the prediction uncertainty of the models, based on spectral feature partition, was relatively low. Our results confirm the effectiveness and stability of UAV hyperspectral and spectral partition-based modeling in developing SPAD spatial distribution estimation models for saline vegetation.