Detail Layer Research Articles

Multimodal Medical Image Fusion Using Hybrid Layer Decomposition With CNN-Based Feature Mapping and Structural Clustering

Multimodal medical image fusion aims to integrate multisensory medical data into a single image for better prediction of the diagnostic details. This article presents a feature-level multimodal medical image fusion with the use of a two-scale ℓ <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sub> - ℓ <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0</sub> hybrid layer decomposition scheme to maximize the structural details with the suppression of significant noise and artifacts. The proposed method uses a convolutional neural network (CNN) with consistency verification and structural patch clustering (fuzzy c-means-based) for the decomposed base and detail layer fusion, respectively. At first, a color space transform is used to separate the luminance and chrominance components extracted from both the source images and the luminance part of each image is decomposed. In the second step, a pretrained CNN model is used to extract the prominent features from each decomposed base layer component. For the output feature map, a regional energy-based activity measure is computed and generated a fusion score, which is refined in the consistency verification step to optimize the weight map for fusing the base layers. The two-scale detail layers are merged by the clustering-based prelearned dictionary for efficiently mapping of structural details of the layers. The color components associated with both the images are also combined using a color saliency measure. Finally, the fused base layer, detail layers, and color components are merged to get the resultant fused image. The experimental results justify the superiority of the proposed approach in terms of both subjective and objective assessments. The performance of the proposed scheme is tested on a large multimodal medical data set of MR-SPECT, MR-PET, and CT-MR neurological images. Furthermore, the proposed approach shows better results than the other state-of-the-art approaches with improved fusion quality and computational performances.

Cardinal orientation and melting temperature effects for PCM-enhanced light-walls in different climates

The use of Phase Change Materials (PCMs) in buildings, aimed at peak shavings and energy savings, has a long history and many applications. In the last decade, the use of PCMs embedded in thermal insulation layers for building envelopes become an interesting option for lightweight structures.Nevertheless, the possibility of profitably exploiting the thermal properties of PCMs is strongly affected by the harmonised combination of the parameters that determine the thermal consumption of the building.In order to analyse the effects of a PCM insulation layer in detail, this paper presents a complete dynamic model of a typical lightweight wall, containing a layer of PCMs, using a thermal model which also considers the hysteresis of specific heats. Five different cases were evaluated to study the effect of PCM positioning on energy utilisation in two different climates (Würzburg, Germany and Denver, USA), considering the effect of solar radiation on differently oriented walls, in summer.Results confirmed that the strong non-linearity of the PCMs’ behaviour and of the weather boundary condition deeply affect the thermal performance of the building, being the main reason for this effect an increase of the thermal capacity of the PCM layer and the external temperature.

Efficient and Adaptive Tone Mapping Algorithm Based on Guided Image Filter

The recognition rate of computer vision algorithms is highly dependent on the image quality. To enhance the visual quality of the images captured under high-dynamic range (HDR) scenes, we propose an efficient and adaptive tone mapping algorithm based on guided image filter (GIF). The HDR image is compressed adaptively according to its average luminance. Then we decompose it into a base layer and a detail layer using the guided image filter. We improve the base layer and enhance the detail layer simultaneously, and combine the two layers to get the final low-dynamic range (LDR) image. Since the parameters are linked with image statistics, they adaptively fit to various kinds of images. The objective evaluation results on HDR image sets demonstrate the superiority of our proposed algorithm. Meanwhile, the result of our algorithm can reduce the halo artifacts and preserve more detail by subjective observation.

Structural path and decomposition analysis of aggregate embodied energy and emission intensities

Aggregate energy and emission intensities have respectively been widely used to measure the overall performance of energy consumption and environmental pollution from the production perspective. Recently, Su and Ang (2017) propose the aggregate embodied intensity (AEI) indicator, defined as the ratio of embodied energy (or emissions) to embodied value added, to analyze the relationship between energy (or emissions) and value added or GDP from the demand perspective using the input-output (I-O) framework. Besides I-O analysis, structural path analysis (SPA) can be used to split the I-O analysis results into different layers to extract the important paths in terms of energy consumption and the resulting emissions. This paper incorporates the SPA technique with the AEI indicators and structural decomposition analysis (SDA) technique in the context of energy and emission studies. An empirical study using China's 2007 and 2012 datasets is presented to illustrate the AEI at the detailed transmission layers, show their relationships with the AEI indicators at different levels, and further investigate the driving forces to the changes of these AEI indicators. The proposed multi-level AEI framework can also be applied to other indicators and extended to multi-country/region analysis.

Infrared and visible image fusion algorithm based on saliency detection and adaptive double-channel spiking cortical model

Based on the saliency detection which effectively extracts the saliency information of the image and the adaptive double-channel spiking cortical model (ADSCM) which has the advantages of global coupling, pulse synchronization, less parameters and high computational efficiency, which can well process information in dark areas, we proposed a new infrared and visible image fusion algorithm. Firstly, the infrared and visible images are decomposed into the base layer and the detail layers through the rolling guidance filter (RGF). Secondly, the base layer is fused with an improved saliency detection method, and the detail layers are fused by a directional gradient sum and an improved ADSCM. The experimental results show that the proposed method can also fully extract the target information of the infrared image while preserving the visible background information, in which the fusion effect is better than the traditional fusion algorithms.

Dielectrical Properties of Living Epidermis and Dermis in the Frequency Range from 1 kHz to 1 MHz.

We determine the in-vivo dielectric properties—resistivity and relative permittivity—of living epidermis and dermis of human skin soaked with a physiological saline solution for one minute between 1 kHz and 1 MHz. This is done by fitting approximate analytical solutions of a mechanistic model for the transport of charges in these layers to a training set comprising impedance measurements at two depth settings on stripped skin on the volar forearm of 24 young subjects. Here, the depth settings are obtained by varying the voltage at a second inject on the electrical-impedance-spectroscopy probe. The model and the dielectric properties are validated with a test set for a third depth setting with overall good agreement. In addition, the means and standard deviations of the thicknesses of living epidermis and dermis are estimated from a literature review as 61±7 μm and 1.0±0.2 mm respectively. Furthermore, extensions to resolve the skin layers in more detail are suggested.

In Vivo Measurement of Plantar Tissue Characteristics and Its Indication for Foot Modeling.

Plantar heel pain is one of the most common musculoskeletal disorders and generally causing long term discomfort of the patients. The objective of the present study is to combine in vivo experimental measurements and finite element modelling of the foot to investigate the influences of stiffness and thickness variation of individual plantar tissues especially the heel pad on deformation behaviours of the human foot. The stiffness and thickness variance of individuals were measured through supersonic shear wave elastography considering detailed heel pad layers refered to in literature as: dermis, stiffer micro-chamber layer, softer macro-chamber layer. A corresponding foot model with separated heel pad layers was established and used to a sensitivity analysis related to the variance of above-mentioned tissue characteristics. The experimental results show that the average stiffness of the micro-chamber layer ranged from 24.7 (SD 2.4) kPa to 18.8 (SD 3.5) kPa with the age group increasing from 20-29years old to 60-69years old, while the average macro-chamber stiffness is 10.6 (SD 1.5) kPa that appears to slightly decrease with the increasing age. Both plantar soft tissue stiffness and thickness of male were generally larger than that of female. The numerical simulation results show that the variance of heel pad strain level can reach 27.5% due to the effects of stiffness and thickness change of the plantar tissues. Their influences on the calcaneus stress and plantar pressure were also significant. This indicates that the most appreciate way to establish a personalized foot model needs to consider the difference of both individual foot anatomic geometry and plantar soft tissue material properties.

A new detail enhancement method for high dynamic range infrared image

Due to the width limitation of 8-bit data used in traditional display device, the local details and contrast of the infrared image are compressed. To cope with these limitations, a novel infrared image enhancement method based on CLAHE and multi-scale approach of DoG is proposed. First, the guided image filter is utilized to separate the base layer and detail layer of the image. Then, the improved CLAHE is used to adjust the base layer global contrast and the multi-scale detail boosting is used to enhance the detail features; for the detail layer, multi-scale fast median filter is used for noise reduction. Finally, the two layers are fused according to the weight of the detail layers’ noise level. Comparing with the real-time enhancement methods both in qualitative and quantitative aspects, our proposed method shows the better performance in experiment results.

Corrosion of Carbon Free and Bonded Refractories for Application in Steel Ingot Casting: An Approach for Improving Steel Quality

Impurities and resulting inclusions are an issue when processing higher amounts of scrap during steel making. To increase the recycling rate, the removal of impurities from the scrap in form of inclusions is of great interest. In previous studies was found that inclusions attach primarily on carbon containing refractories, especially if on their surface an interfacial layer (1–3 µm thickness) was formed in-situ. This study investigates the formation mechanism of this in-situ layer in detail by application of computer tomography (CT) measurements on two scales. The large scale CT scans visualized the general appearance whereas the small scale measurement regarded the in-situ formed layer and the attached inclusions in detail. Based on these measurements, previous results and a literature review it was concluded that the layer formed mainly due to carbothermally reduced impurities which moved to the decarburized surface of the refractory in gaseous form and enhanced sintering of the surface region to develop the layer.

A hierarchical approach to analyzing knowledge integration between two fields—a case study on medical informatics and computer science

As a driver in modern science, interdisciplinary research has attracted a lot of attention. Major foci are laid on exploring the relations of multiple involved disciplines as well as the knowledge structure in interdisciplinary field. However, there is still a lack of decomposing the knowledge structure of interdisciplinary field to investigate how knowledge from relevant disciplines is integrated in the field. This study proposes an approach to investigating knowledge integration relationships between two research fields from a perspective of hierarchy. Medical Informatics (MI) and its most relevant field of Computer Science (CS) are chosen in the case study. This study decomposed each keyword network of the two fields into four layers by using the K-core method, then quantified the knowledge integration relationships between different layers of the two fields together. The results present that the MI basic layer shows the strongest knowledge integration with CS, followed by the middle layer, with the detail layer the weakest. And all MI layers have the greatest breadth and strength of knowledge integration with the CS middle layer, followed by the CS marginal layer and detail layer, but with the CS basic layer the weakest. A time series analysis shows that the integration of new CS knowledge into MI is a gradual process without explosive growth and the path of knowledge integration between the two fields were identified. The proposed approach could be applied to deeply understanding the integration of one discipline knowledge by an interdisciplinary field.

TONE COMPRESSION ALGORITHM FOR HIGH DYNAMIC RANGE MEDICAL IMAGES

Abstract. We propose here an HDR compression method for medical images based on a windowing operator, an adaptive tone mapping operator, and the probabilistic normal-gamma model. First, we use the windowing operator based on a structural fidelity measure for optimal visualization of the input HDR medical image. Then, we transform the windowed image to the logarithm domain and split it into base and detail layers with the help of the probabilistic normal-gamma model. Base and detail layers are used to make the tone map with help the adaptive tone mapping operator. Finally, the tone mapping result is the LDR image. The proposed method has comparable quality and low computation time compared to other tone mapping operators.

Super-resolution reconstruction of neonatal brain magnetic resonance images via residual structured sparse representation.

Magnetic resonance images of neonates, compared with toddlers, exhibit lower signal-to-noise ratio and spatial resolution. In this paper, we propose a novel method for super-resolution reconstruction of neonate images with the help of toddler images, using residual-structured sparse representation with convex regularization. Specifically, we introduce a two-layer image representation, consisting of a base layer and a detail layer, to cater to signal variation across scanners and sites. The base layer consists of the smoothed version of the image obtained via Gaussian filtering. The detail layer is the difference between the original image and the base layer. High-frequency details in the detail layer are borrowed across subjects for super-resolution reconstruction. Experimental results on T1 and T2 images demonstrate that the proposed algorithm can recover fine anatomical structures, and generally outperform the state-of-the-art methods both qualitatively and quantitatively.

Resolution enhancement algorithm based on infrared digital holography imaging through flame

In recent years, the use of new technologies combining infrared thermal imaging and digital holographic imaging to observe the targets in the fire field has become a current research focus. In theory, flame and smoke have no effect on long-wavelength infrared digital holography, but in the real fire environment, large particles of dust from the combustion will interfere with the light path, seriously increasing the reconstruction noise of the hologram. This paper proposes a new image processing algorithm to suppress the noise of infrared digital holographic reconstruction. The algorithm uses a bilateral filter combined with the Laplacian pyramid algorithm to separate the details and energy layers of the holographic reconstructed image, filters the detail layer, and then superimposes the separated layers back into the reconstructed image by the inverse Laplacian pyramid algorithm. Therefore, the resolution of the reconstructed image is improved, and the simulation of the fire field environment proves that the algorithm has a significant effect on improving the resolution of the reconstructed image of the infrared digital hologram.

Multi-focus image fusion with random walks and guided filters

Multi-focus image fusion technique is able to help obtaining an all-focused image, which is advantage to human vision and image processing. In this paper, a novel multi-focus image fusion method is proposed based on random walk and guided filter. In the proposed method, the decomposition function and the optimizing function of random walk are used in multi-focus image fusion. And the random walk is also utilized for weight maps directly. The advantages of random walk and guided filter in image fusion are fully utilized by regulating proportional coefficients artificially. The proposed method concludes six steps: first, decomposing source images into detail layers and base layers with random walk; second, the random walk is used for weight maps directly and the guided filter is used as smoothing filters to get the streamlined weight maps of the detail layers and the base layers, respectively; third, the weight maps of the detail layers and the base layers are acquired by summing the initializing weight maps in different proportions; and then, the final weight maps of the detail layers are acquired using random walk for optimizing. After that, the fused detail layer and base layer are obtained by weighted average of detail layers and base layers, singly. Finally, the fused image is gained by summing up the fused base layer and the fused detail layer. Experiments demonstrate that the proposed method outperforms many other approaches in both subjective and objective assessments.

Data-driven two-layer visual dictionary structure learning

An important issue in statistical modeling is to determine the complexity of the model based on the scale of data so as to effectively mitigate the model’s overfitting problems without big data. We adopt a data-driven approach to automatically determine the number of components of the model. In order to better extract robust features, we propose a framework of data-driven two-layer structure visual dictionary learning (DTSVDL). It works by dividing the visual dictionary structure learning into two levels: the attribute layer and the detail layer. In the attribute layer, the attributes of the image dataset are learned, and these attributes are obtained by a data-driven Bayesian nonparametric model. Then, in the detail layer, the detailed information over attributes is further explored and refined, and the attributes are weighted by the number of effective observations associated with each attribute. Our proposed approach has three main advantages: (1) the two-layer structure makes our building visual dictionary be more expressive; (2) the number of components in the attribute layer can be determined automatically from the data; (3) the components are automatically determined based on the scale of visual words; therefore, our model can well mitigate the overfitting problem. In addition, by comparing with stacked autoencoders, stacked denoising autoencoders, LeNet-5, speeded-up robust features, and pretrained deep learning model ImageNet-VGG-F algorithms, we find that our approach achieves satisfactory image categorization results on two benchmark datasets. Specifically, higher categorization performance is achieved than by the classical approaches on 15 scene categories and action datasets. We conclude that the resulting DTSVDL possesses a good generality derived from attribute information as well as an excellent distinction derived from detailed information. In other words, the visual dictionary learned by our algorithm is more expressive and discriminatory.

Example based facial aging simulation via facial detail transfer

The key problem of the facial aging simulation is protecting the identity information during the texture synthesizing procedure. In this paper, we address the aging simulation problem by using the facial detail transfer method, which provides a fundamental view of the aging simulation problem and shows its superiority in the small dataset. The proposed method decomposes the facial image into several layers, namely, structure layer, detail layer and color layer. The structure layer protects the identity information of the test image, while the detail layer contains the age information from the selected template image. We transfer the detail layer from the example images to the test images to generate the aging effects on the synthesized images. The images near the target age are also considered in the proposed method to further refine the synthesized results in the aging sequence. Experimental results conducted on the FG-NET database, Morph database and the IRIP-Aging database show that the proposed method can not only protect the identity information and the age information, but also generate more smooth aging sequence, which demonstrates the effectiveness of the proposed method.

Fast local SAR image despeckling by edge-avoiding wavelets

In this paper, a local despeckling algorithm based on a trous version of edge-avoiding wavelets for synthetic aperture radar images is proposed. Edge-avoiding wavelets provide the decomposition of the input image into several detail layers and a residual image which corresponds to the despeckled result by means of weighted a trous wavelet transform. The weight function is similar to the range kernel of the well-known bilateral filter and is determined by enhancing evaluation indexes. Visual and quantitative comparisons validate the performance of the proposed method in terms of both speckle reduction and edge preservation ability.

Learning-Based Low-Complexity Reverse Tone Mapping With Linear Mapping

Although high dynamic range (HDR) display has become popular recently, the legacy content such as standard dynamic range (SDR) video is still in service and needs to be properly converted on HDR display devices. Therefore, it is desirable for HDR TV sets to have the capability of automatically converting input SDR video into HDR video, which is called reverse tone mapping (RTM). In this paper, we propose a novel learning-based low-complexity RTM scheme that not only expands the suppressed dynamic ranges (DR) of the SDR videos (or images), but also effectively restores lost detail in the SDR videos. Most existing conventional RTM schemes have focused on how to expand the DR of global contrast, resulting in limitations in recovering lost detail of SDR videos. On the other hand, the recent convolutional neural network-based approaches show promising results, but they are too complex to be applied on the users’ devices in practice. In this paper, our learning-based RTM scheme is computationally simple but effective in recovering lost detail. To learn the SDR-to-HDR relation, training “SDR-HDR” images are first separated into their base layer components and detail layer components by applying a guided filter. The detail layer components of the “SDR-HDR” pairs are used to train the SDR-to-HDR mapping. The mapping matrices are computed based on kernel ridge regression. In the meantime, the global contrast of the base layers is expanded by a nonlinear function that suppresses darker regions and amplifies brighter regions to fit the full DR of a target HDR display. To verify the effectiveness of our learning-based RTM scheme, we performed subjective quality assessment for images and videos. The experimental results show that our RTM scheme outperforms the existing RTM scheme with the successful restoration of lost detail in SDR images.

Fusion of infrared and visible images with Gaussian smoothness and joint bilateral filtering iteration decomposition

Edge‐preserving filters have been applied to Multi‐Scale Decomposition (MSD) for fusion of infrared and visible images. Traditional edge‐preserving MSDs may hardly make satisfied structural separation from details to cause fusion performance degradation. To suppress this challenge, the authors propose a novel fusion of infrared and visible images with Gaussian smoothness and joint bilateral filtering iteration decomposition (MSD‐Iteration). This method consists of three steps. First, source images are decomposed by the Gaussian smoothness and joint bilateral filtering iteration. The implementation includes the fine‐scale detail removal with Gaussian filtering, edge and structure extraction with joint bilateral filtering iteration, and detail obtaining at multi‐scales. The decomposition has edge‐preserving and scale‐aware properties to improve detail acquisition. Second, rules are designed to conduct the layer combination. For the rule of base layers, saliency maps are constructed by Laplacian and Gaussian low‐pass filters to calculate initial weight maps. A guided filter is further applied to determine final weight maps for the combination. Meanwhile, they use the regional average energy weighting to obtain decision maps at multi‐scales by constructing intensity deviation to combine detail layers. Third, they implement the reconstruction with the combined layers. Sufficient experiments are presented to evaluate MSD‐Iteration, and experimental results validate the superiority of the authors’ method.

Ground penetrating Radar Clutter Removal via 1D Fast Sub band Decomposition

Target detection performance in ground-penetrating radar (GPR) deteriorates highly in the presence of clutter. Multi-scale (wavelet transform) or the recently proposed multi-scale and multi-directional decomposition based methods can efficiently remove the clutter, however they have high computational complexity. In this paper, we propose a new multi-scale method which requires only 1D fast subband decomposition of the rows of the GPR image. The resulting detail layers directly provide the clutter-free target component of the GPR image. The proposed method is compared to the state-of-art clutter removal methods both visually and quantitatively using a realistic simulated dataset which is constructed by the gprMax simulation software. The results show that the proposed 1D subband decomposition scheme approximates the classical 2D wavelet decomposition successfully and even presents a performance increase as well as a complexity decrease for fast decomposition methods based on lifting wavelet transform and a trous wavelet transform.