High Frequency Subbands Research Articles

Regularity-Constrained Fast Sine Transforms

This letter proposes a fast implementation of the regularity-constrained discrete sine transform (R-DST). The original DST \textit{leaks} the lowest frequency (DC: direct current) components of signals into high frequency (AC: alternating current) subbands. This property is not desired in many applications, particularly image processing, since most of the frequency components in natural images concentrate in DC subband. The characteristic of filter banks whereby they do not leak DC components into the AC subbands is called \textit{regularity}. While an R-DST has been proposed, it has no fast implementation because of the singular value decomposition (SVD) in its internal algorithm. In contrast, the proposed regularity-constrained fast sine transform (R-FST) is obtained by just appending a regularity constraint matrix as a postprocessing of the original DST. When the DST size is $M\times M$ ($M=2^\ell$, $\ell\in\mathbb{N}_{\geq 1}$), the regularity constraint matrix is constructed from only $M/2-1$ rotation matrices with the angles derived from the output of the DST for the constant-valued signal (i.e., the DC signal). Since it does not require SVD, the computation is simpler and faster than the R-DST while keeping all of its beneficial properties. An image processing example shows that the R-FST has fine frequency selectivity with no DC leakage and higher coding gain than the original DST. Also, in the case of $M=8$, the R-FST saved approximately $0.126$ seconds in a 2-D transformation of $512\times 512$ signals compared with the R-DST because of fewer extra operations.

Deep Shearlet Network for Change Detection in SAR Images

Convolutional neural networks (CNN) can extract shift-invariant features, and have been widely applied in change detection task. However, common CNN lacks noise robustness and needs supervised data, to alleviate these problems, in this paper, we propose a novel deep shearlet network (ShearNet) for change detection in SAR images. In the network, a shearlet denoising layer (SDL) is designed to enhance the representation ability of common CNN. In SDL, feature maps are decomposed into subband coefficients by shearlet transform (ST). Due to optimal sparse representation property and highly direction sensitivity of ST, the network can capture important geometric information. Then, hard-threshold shrinkage is applied to high frequency subbands to drop small coefficients that are most likely to be noise, so that reduce the effect of noise. Finally, ShearNet is trained by introducing a noise-robust loss with noisy labels. The noisy labels are obtained by deep clustering that shows more robustness than existing preclassification methods. This fine-tuning process novelly follows the paradigm of learning from noisy labels to aside the difficulty of precisely labeling samples. Our experimental results on multiple real SAR datasets show that ShearNet can boost accuracy, and have better applicability for change detection in SAR images. The source code is available at https://github.com/yizhilanmaodhh/ShearNet.

An Image Steganography Algorithm Based on PSO and IWT for Underwater Acoustic Communication

To achieve fast and secure information steganography in underwater acoustic channel scenarios, this paper proposes a PSO (Particle Swarm Optimization) and IWT (integer wavelet transform) based image steganography algorithm for underwater acoustic communication. Firstly, the underwater image to be embedded is interleaved and packed to generate the secret information code-stream to improve the detection and recovery performance of transmission errors. Secondly, the cover image is divided into non-overlapping blocks and IWT is applied to each block, which can save PSO search space and improve the optimal solution performance. Thirdly, the embedding position matrix of the high frequency sub-band coefficients is chaotic scrambled, and PSO is used to select the optimal key to obtain the optimal matching position for the secret information to produce the best stego-image quality. Finally, transmission error detection and recovery measures are applied to the extracted secret information to improve the reconstruction quality in underwater acoustic channel scenarios. Experimental results show that the proposed algorithm outperforms existing approaches in terms of information embedding capacity and steganographic visual quality, meeting the requirements of safety and real-time performance of underwater communication. Moreover, it also has the ability of error detection and recovery for harsh underwater acoustic channel.

Medical Image Enhancement through Dual Tree Complex Wavelet Transform and Multi-resolution Gabor Filter

Medical image processing is essential in the field of clinical medicine for analysis and diagnosis of diseases. Most of these medical images are acquired under low illumination and noisy environment. Hence it produces a bottleneck in the processing of medical images through computer vision system. Due to tremendous details and complexity of the medical images, image enhancement is challenging and difficult task. In this paper, medical image enhancement through dual tree complex wavelet transform (DTCWT) and multi-resolution Gabor filter is proposed. After the medical images are acquired using standard techniques, DTCWT is applied that decomposes the gray scale image into high and low frequency subbands. Further DTCWT is applied twice on low frequency subbands to decompose into its high and low frequency subbands. Then multiresolution Gabor filter is applied to decomposed six high frequency subands and two low frequency subband to obtain enhancement in multiple directions. Finally inverse DTCWT is applied to obtain enhanced medical image. The experimental results demonstrate that the approach based on DTCWT and multi-resolution Gabor filter can produce better image enhancement as compared with the conventional and latest methods. The proposed medical image enhancement technique can be applied in analysis and diagnosis of diseases through computer vision system.

Perceptual Fusion of Electronic Chart and Marine Radar Image

Electronic charts and marine radars are indispensable equipment in ship navigation systems, and the fusion display of these two parts ensures that the vessel can display dangerous moving targets and various obstacles on the sea. To reduce the noise interference caused by external factors and hardware, a novel radar image denoising algorithm using the concept of Generative Adversarial Network (GAN) using Wasserstein distance is proposed. GAN focuses on transferring the image noise distribution between strong and weak noise, while the perceptual loss approach is to suppress the noise by comparing the perceptual characteristics of the output after denoising. Afterwards, an image registration method based on image transformation is proposed to eliminate the imaging difference between the radar image and chart image, in which the visual attribute transfer approach is used to transform images. Finally, the sparse theory is used to process the high frequency and low frequency subband coefficients of the detection image obtained by the fast Fourier transform in parallel to realizing the image fusion. The results show that the fused contour has a high consistency, fast training speed and short registration time.

Stereo Video Quality Metric Based on Multi-Dimensional Analysis.

Stereo video has been widely applied in various video systems in recent years. Therefore, objective stereo video quality metric (SVQM) is highly necessary for improving the watching experience. However, due to the high dimensional data in stereo video, existing metrics have some defects in accuracy and robustness. Based on the characteristics of stereo video, this paper considers the coexistence and interaction of multi-dimensional information in stereo video and proposes an SVQM based on multi-dimensional analysis (MDA-SVQM). Specifically, a temporal-view joint decomposition (TVJD) model is established by analyzing and comparing correlation in different dimensions and adaptively decomposes stereo group of frames (sGoF) into different subbands. Then, according to the generation mechanism and physical meaning of each subband, histogram-based and LOID-based features are extracted for high and low frequency subband, respectively, and sGoF quality is obtained by regression. Finally, the weight of each sGoF is calculated by spatial-temporal energy weighting (STEW) model, and final stereo video quality is obtained by weighted summation of all sGoF qualities. Experiments on two stereo video databases demonstrate that TVJD and STEW adopted in MDA-SVQM are convincible, and the overall performance of MDA-SVQM is better than several existing SVQMs.

No-reference stereoscopic image quality assessment using quaternion wavelet transform and heterogeneous ensemble learning

As the demand for high-quality stereo images has grown in recent years, stereoscopic image quality assessment (SIQA) has become an important research area in modern image processing technology.In this paper, we propose a no-reference stereoscopic image quality assessment (NR-SIQA) model using heterogeneous ensemble learning ‘quality-aware’ features from luminance image, chrominance image, disparity and cyclopean images via quaternion wavelet transform (QWT). Firstly, luminance image and chrominance image are generated by CIELAB color space as monocular perception, and the novel disparity and cyclopean images are utilized to complement with monocular information. Then, a number of ‘quality-aware’ features in the quaternion wavelet domain are discovered, including entropy, texture features, energy features, energy differences features and MSCN coefficients of high frequency sub-band. Finally, a heterogeneous ensemble model via support vector regression (SVR) & extreme learning machine (ELM) & random forest (RF) is proposed to predict quality score, and bootstrap sampling and rotated feature space are used to increase the diversity of data distribution. Comparing with the state-of-the-art NR-SIQA models, experimental results on four public databases prove the accuracy and robustness of the proposed model.

Privacy-preserving reversible data hiding based on quad-tree block encoding and integer wavelet transform

Applications on the cloud server have matured, and protecting the privacy of the content owner has attracted more attention. Privacy-Preserving Reversible data hiding (PP-RDH) is an efficient technique for embedding additional data into an encrypted image. In this paper, we propose a privacy-preserving reversible data hiding scheme using the quad-tree partition and Integer Wavelet Transform (IWT) techniques. Our scheme focuses on improving the embedding rate and quality of the recovered image when a 2 × 2-sized, block-based image encryption method is applied to ensure relative higher security. On this basis, the IWT technique transforms the encrypted image, and coefficients in three high frequency subbands are converted into 8-bit binary system. Then, the quad-tree partition technique encodes each 8 × 8-sized coefficient block, since there are many zeroes in the front bit planes. The experimental results indicated that our proposed scheme significantly improved the embedding rate, and guaranteed lossless image recovery and data extraction.

Degraded Image Enhancement Using Dual-Domain-Adaptive Wavelet and Improved Fuzzy Transform

A novel enhancement algorithm for degraded image using dual-domain-adaptive wavelet and improved fuzzy transform is proposed, aiming at the problem of surveillance videos degradation caused by the complex lighting conditions underground coal mine. Firstly, the dual-domain filtering (DDF) is used to decompose the image into base image and detail image, and the contrast limited adaptive histogram enhancement (CLAHE) is adopted to adjust the overall brightness and contrast of the base image. Then, the discrete wavelet transform (DWT) is utilized to obtain the low frequency sub-band (LFS) and high frequency sub-band (HFS). Next, the wavelet shrinkage threshold is applied to calculate the wavelet threshold corresponding to the HFS at different scales. Meanwhile, a new Garrate threshold function that introduces adjustment factor and enhancement coefficient is designed to adaptively de-noise and enhance the HFS coefficients, and the Gamma function is employed to correct the LFS coefficients. Finally, the PAL fuzzy enhancement operator is improved and used to perform contrast enhancement and highlight area suppression on the reconstructed image to obtain an enhanced image. Experimental results show that the proposed algorithm can not only significantly improve the overall brightness and contrast of the degraded image but also suppresses the noise of dust and spray and enhances the image details. Compared with the similar algorithms of STFE, GTFE, CLAHE, SSR, MSR, DGR, and MSWT algorithms, the indicator values of comprehensive performance of the proposed algorithm are increased by 205%, 195%, 200%, 185%, 185%, 85%, 140%, and 215%, respectively. Moreover, compared with the other seven algorithms, the proposed algorithm has strong robustness and is more suitable for image enhancement in different mine environments.

Image Retrieval using the Improved Double Density Contourlet Transform

A new image retrieval algorithm based on the improved double density contourlet transform(DDCT) is proposed in this paper. The improved double density contourlet transform have the property of shift invariance and non-Gaussian for the high frequency sub-bands. In order to obtain the high frequency texture feature, we design the quantization histograms using the high frequency direction sub-bands. The 0/1 quantization is used to deal with the energy matrix of high frequency sub-bands. A texture point is defined as an 8-D vector by integrating different channel values from high frequency sub-bands and the quantization histograms are computed. In order to get the texture-spatial features, the local bina ry pattern is used to describe the texture feature of low freq uency sub-band. Then the quantization histograms and the local binary pattern (LBP) can be used to denote the texture features of the image. Experiments show that the proposed algorithm using the improved double density contourlet transform outperforms the SD algorithm based on the contourlet transform in the natural image retrieval.

Contrast Enhancement Technique using Discrete Wavelet Transform with Just Noticeable Difference Model for 3D Stereoscopic Degraded Video

The Video Technologies for Medical, cultural, and social activities prefer 3D visual data rendering and processing. So 3D videos are captured by any capturing devices, like the digital cameras are not acceptable all the time due to the lack of capturing devices or indecent illumination or due to poor weather surroundings like Low light, rain, fog, mist, etc. reduces the contrast, thus the videos get degraded. 3D video contrast enhancement technique is an essential process for upgrading the quality and information content in the videos. The proposed work employs a discrete wavelet transform based enhancement technique with Jut noticeable difference model to improve the video frames and it is simple and computationally inexpensive. The application of DWT results in the Low and High-frequency sub-bands. The low-frequency components that contain the greatest amount of the information are improved using weighted threshold histogram equalization(WTHE) with the JND model algorithm while the high-frequency sub-bands are distortions and highly affected by noise. The Gaussian high pass filter is applied to each high-frequency sub-bands to remove the noise. Besides, enhancement gain control and luminance preservation are used to acquire the enhanced output video. At the end check the quality of the degraded video frame, the presented work is implemented in MATLAB 2018a and evaluated using objective parameters. Experimental results show that the proposed method can generate better and agreeable results than 2D videos.

Medical image fusion based on convolutional neural networks and non-subsampled contourlet transform

Although many powerful convolutional neural networks (CNN) have been applied to various image processing fields, due to the lack of datasets for network training and the significant different intensities of diverse multi-modal source images at the same location, CNN cannot be directly used for the field of medical image fusion (MIF), which is a major problem and limits the development of this field. In this article, a novel multimodal medical image fusion method based on non-subsampled contourlet transform (NSCT) and CNN is presented. The proposed algorithm not only solves this problem, but also exploits the advantages of both NSCT and CNN to obtain better fusion results. In the proposed algorithm, source multi-modality images are decomposed into low and high frequency subbands. For high frequency subbands, a new perceptual high frequency CNN (PHF-CNN), which is trained in the frequency domain, is designed as an adaptive fusion rule. In the matter of the low frequency subband, two result maps are adopted to generate the decision map. Finally, fused frequency subbands are integrated by the inverse NSCT. To verify the effectiveness of the proposed algorithm, ten state-of-the-art MIF algorithms are selected as comparative algorithms. Subjective evaluations by five doctors as well as objective evaluations by seven image quality metrics, demonstrate that the proposed algorithm is superior to the other comparative algorithms in terms of fusing multimodal medical images.

Subband Adaptive Image Deblocking Using Wavelet Based Convolutional Neural Networks

In this paper, we propose subband adaptive image deblocking using wavelet based convolutional neural networks (CNNs). We build wavelet based CNNs for image deblocking to achieve subband adaptive reconstruction. First, we perform subband adaptive processing after the discrete wavelet transform (DWT) on the input image. For low frequency subband (LL), we use a simple and effective shallow CNN to restore the low frequency component, while for high frequency subbands (LH, HL, and HH) we utilize multi-kernel convolution to capture multiscale features and restore sparse high frequency components. Then, we conduct mixed convolution of dilated convolution and standard convolution to expand the receptive field while introducing channel and spatial attentions to adjust the proportion of different subbands and spatial coordinates. Various experiments on Classic5 and LIVE1 datasets show that the proposed method successfully recovers sharp edges and clear textures in highly compressed images while removing compression artifacts such as blocking and banding. Moreover, the proposed method achieves comparable state-of-the-art performance on compression artifact removal in terms of both visual quality and quantitative measurements.

A Robust Medical Images Watermarking Using FDCuT-DCT-SVD

One way to prevent image duplication is by applying watermarking techniques. In this work, the watermarking process is applied to medical images using the Fast Discrete Curvelet Transforms (FDCuT), Discrete Cosine Transform (DCT), and Singular Value Decomposition (SVD) methods. The medical image of the host is transformed using FDCuT so that three subbands are obtained. High Frequency (HF) subband selected for DCT and SVD applications. Meanwhile, SVD was also applied to the watermark image. The singular value on the host image is exchanged with the singular value on the watermark. Insertion of tears by exchanging singular values does not cause the quality of medical images to decrease significantly. The experimental results prove that the proposed FDCuT-DCT-SVD algorithm produces good imperceptibility. The proposed algorithm is also resistant to various types of attacks, including JPEG compression, noise enhancement attacks, filtering attacks, and other common attacks.

Brain medical image fusion scheme based on shuffled frog‐leaping algorithm and adaptive pulse‐coupled neural network

Aiming at the problems of low contrast and blurred edge textures in medical image fusion, a new fusion scheme in non-subsampled contourlet transform (NSCT) domain is proposed to improve the quality of fused brain images which is based on pulse-coupled neural network (PCNN) and shuffled frog-leaping algorithm (SFLA). First, the source images are decomposed into low-frequency (LF) and high-frequency (HF) subbands using NSCT; if one of the source images is multicolour, then hue, saturation and brightness (HSI) transform is needed first. Second, different PCNN fusion rules are designed for LF and HF subbands according to their features, respectively. Parameters including decay time constants and amplification factors are optimised by SFLA. Finally, the fused image is reconstructed by inverse NSCT; and if necessary, an inverse HSI transform is needed. Visual and quantitative analysis of experimental results show that the fused image preserves more information of the source images, and the ability of edge retention is strong. The scheme has prominent advantages in mutual information and QAB/F for multimodal brain images, including MRI-PET, MRI-SPECT, and CT-MRI, which proves that it can obtain better visual effect and have strong robustness as well as wide applications.

The Rolandic operculum generates different semiologies in insulo-opercular and temporal lobe epilepsies

PurposeThe role of the Rolandic operculum in in mesial temporal lobe epilepsy (MTLE) is to produce oroalimentary automatisms (OAAs). In insulo-opercular epilepsy (IOE), the Rolandic operculum may produce perioral muscle clonic or tonic movements or contractions. This paper aims to confirm the symptomatogenic zone of facial symptoms in IOE and to explain this phenomenon. MethodsA total of 45 IOE patients and 15 MTLE patients were analyzed. The patients with IOE were divided into facial (+) and (-) groups according to the facial symptoms. The interictal positron emission tomography (PET) data were compared among groups. Furthermore, electroclinical correlation, functional connectivity and energy ratio (ER) were analyzed with stereo-electroencephalography (SEEG). ResultsIntergroup PET differences were observed mainly in the Rolandic operculum. Electroclinical correlation showed that the Rolandic operculum was the only brain area showing any correlations. Compared with the facial (-) group, the facial (+) group showed stronger functional connectivity and a higher ER in the alpha 1, alpha 2 and beta sub-bands. In the Rolandic operculum, compared with those of the MTLE group, the h2 and ER of the facial (+) group were higher in the high frequency sub-bands. Intergroup comparison of the ER in the seizure onset zones (SOZ) showed no significant difference. SignificanceThe symptomatogenic zone of facial symptoms in IOE is the Rolandic operculum. Seizure propagation to the Rolandic operculum generates different semiologies because of the different synchronization frequencies and energies of the sub-bands depending on the site of seizure origin. This may be due to the complex spreading pathway from the SOZ to the symptomatogenic zone.

A wavelet transform-based photon starvation artifacts suppression algorithm in CT imaging

In computed tomography, high attenuation occurs when x-rays pass through a dense region or a long path in the scanning object. In this case, only limited photons reach the detector, which causes photon starvation artifacts. The artifacts usually appear as streaks along the directions with high attenuation. It might lower the discrimination of minor structures and lead to misdiagnosis. Applying a local filter to the projection data adaptively is a common solution, however, if the parameters of projection-based filter are not well selected, new artifacts and noise might appear in the final image. In this paper, a post image processing technique was developed to suppress the photon starvation streak artifacts. Based on the directional characteristics of streaks, a semi-adaptive anisotropic diffusion filter was applied to the high frequency sub-bands after wavelet transformation (WASA). Qualitative and quantitative experiments were performed on phantom data and clinical data to prove the effectiveness of this method for photon starvation artifact suppression.

A novel framework based on wavelet transform and principal component for face recognition under varying illumination

One of the most primitive problems in face recognition is illumination variation, which is of significance for image processing and pattern recognition. Existing studies concentrate on the wavelet transform (WT) without elaborately considering high-frequency information and most do not efficiently tackle a computation explosion of facial dimensions in classification. Therefore, a novel framework based on wavelet transform and principal component is proposed to improve the accuracy under illumination variation in this paper. In the proposed framework, low-frequency sub-band (LFSB) and high-frequency sub-band (HFSB) images from wavelet transform are simultaneously enhanced and denoised, unlike previous studies that usually cause loss of details due to less consideration of HFSB. For LFSB image, a multiple scale Retinex-based steering kernel is designed to enhance more details, and then an adaptive strategy of gamma correction is developed to automatically expand gray-dynamic range. For HFSB image, a non-local mean filtering is established to suppress the noise and subsequently, more image details are preserved by local of mean of local variance. Moreover, the principal component technique based Fisherface and virtual auxiliary sample strategy is developed in order to overcome the computation explosion of facial dimensions, in which a sample strategy with interpolation mechanism is employed to avoid the complicated singularity and Fisherface analysis is further applied to extract features and dimensionality reduction. In addition, the particle swarm optimization-neural network (PSO-NN) is employed to perform classification in the framework. Experimental results prove that the proposed framework can effectively obtain the robust visual effect under varying illumination and significantly improve the recognition performance in comparison to existing methods.

Rotation invariant angle-density based features for an ice image classification system

One of the natural disasters which cause economic loss and are a serious threat to society are ice covering phenomena for overhead transmission power lines. This paper presents a new method for ice and non-ice image classification to improve ice detection results. The proposed method uses wavelet decomposition to extract robust features, such as those invariant to rotation, scaling and thickness of ice for classification. The proposed approach estimates the average of the high frequency sub-bands for each level. Then it obtains Canny edge components for the average wavelet image at each level. Our approach studies shape of edge components to identify the presence of ice. To achieve this, the proposed method finds the major and minor axes for each edge component, and then draws parallel lines to the major and minor axes over the edge components. For each parallel line to the major and minor axes, the proposed approach further extracts angle and density-based features for pixels that fall on the parallel lines to the major and minor axes. Next, our method selects features from each average wavelet image and further calculates the mean for the feature vectors corresponding to the level, which results in a feature matrix. Finally, the feature matrix is fed to a Multi-Layer Perceptron Neural Network for classifying ice and non-ice images. Experimental results on a diversified dataset and comparative study with an existing method show that the proposed method is useful for accurate ice detection with better accuracy.

Perceptual quality assessment for multimodal medical image fusion

Recent years have witnessed that the multimodal medical image fusion (MMIF) plays critical roles in clinical diagnostics and treatment. Many MMIF algorithms have been proposed to improve the MMIF images quality. The quality of multimodal medical fused images will significantly affect the results of the clinical diagnosis. However, little work has been designed to evaluate the effectiveness of MMIF algorithms and the quality of MMIF images. To this end, this paper presents a perceptual quality assessment method for MMIF. A MMIF image database (MMIFID) is first built to employ the classical MMIF algorithms, and the subjective experiment is conducted to assess the quality of each fused image. Then, a no-reference objective method is proposed for the perceptual quality evaluation of MMIF images,which uses Pulse Coupled Neural Network (PCNN) in Non-subsampled Contourlet Transform (NSCT). A fused image is decomposed by NSCT into low frequency sub-band (LFS) and high frequency sub-band (HFS). It is used to motivate the PCNN processing, and large firing times are employed to measure LFS and HFS. Finally, two components evaluation results are combined to obtain the overall objective quality score. Experimental results based on the MMIFID indicate that our presented method outperforms the existing image fusion quality evaluation metrics, and it provides a satisfactory correlation with subjective scores, which shows effectiveness in the quality assessment of medical fused images.