Color distortion correction without noise boost using multi-resolution analysis

Gram-Schmidt and Ehlers fusion, which are well known for spectral fidelity, are described. Selecting a sandrock mine using SPOT5 as study area, spectral fidelity and high spatial information gain are used to assess the two fusion methods, which are compared with multiplication and Andorr fusion. In the whole, the Gram-Schmidt method is the best, preserving highly the original spectral information, and can provide spectrum control foundation for interpreting mine targets in the complex geology environment. Ehlers method is the second. Then Andorr method is the third, and it has the highest spatial information gain, but high frequency information is enlarged excessively, effecting on the identification of mine exploitation state. The multiplication method is the worst, because it loses the high and low frequency information, which is the most important for mine targets recognition.

This paper describes an approach to image fusion using the wavelet transform. When images are merged in wavelet space, we can process different frequency ranges differently. For example, high frequency information from one image can be combined with lower frequency information from another, for performing edge enhancement. We have built a prototype system that allows experimentation with various wavelet array combination and manipulation methods for image fusion, using a set of basic operations on wavelet frequency blocks. Problems caused by image misregistration and processing artifacts are described. Examples of wavelet fusion results are shown which merge a pair of images from different sensors.

Time–frequency domain based optimization of hedging strategy: Evidence from CSI 500 spot and futures

First Page

Bilateral filtering has been applied in the area of digital image processing widely, but in the high gradient region of the image, bilateral filtering may generate staircase effect. Bilateral filtering can be regarded as one particular form of local mode filtering, according to above analysis, an mixed image de-noising algorithm is proposed based on Gaussian filter and bilateral filtering. First of all, it uses Gaussian filter to filtrate the noise image and get the reference image, then to take both the reference image and noise image as the input for range kernel function of bilateral filter. The reference image can provide the image’s low frequency information, and noise image can provide image’s high frequency information. Through the competitive experiment on both the method in this paper and traditional bilateral filtering, the experimental result showed that the mixed de-noising algorithm can effectively overcome staircase effect, and the filtrated image was more smooth, its textural features was also more close to the original image, and it can achieve higher PSNR value, but the amount of calculation of above two algorithms are basically the same.

In multichannel cochlear implants, low frequency information is delivered to apical cochlear locations while high frequency information is delivered to more basal locations, mimicking the normal acoustic tonotopic organization of the auditory nerves. In clinical practice, little attention has been paid to the distribution of acoustic input across the electrodes of an individual patient that might vary in terms of spacing and absolute tonotopic location. In normal-hearing listeners, Başkent and Shannon (J. Acoust. Soc. Am. 113, 2003) simulated implant signal processing conditions in which the frequency range assigned to the array was systematically made wider or narrower than the simulated stimulation range in the cochlea, resulting in frequency-place compression or expansion, respectively. In general, the best speech recognition was obtained when the input acoustic information was delivered to the matching tonotopic place in the cochlea with least frequency-place distortion. The present study measured phoneme and sentence recognition scores with similar frequency-place manipulations in six Med-El Combi 40+ implant subjects. Stimulation locations were estimated using the Greenwood mapping function based on the estimated electrode insertion depth. Results from frequency-place compression and expansion with implants were similar to simulation results, especially for postlingually deafened subjects, despite the uncertainty in the actual stimulation sites of the auditory nerves. The present study shows that frequency-place mapping is an important factor in implant performance and an individual implant patient's map could be optimized with functional tests using frequency-place manipulations.

Principal component analysis (PCA) is a method of processing high-dimensional feature data, which can be decomposed into a set of unrelated variables called principal components. Digital holographic microscopy (DHM) is a powerful tool in the biomedical imaging for recording the amplitude and phase information of object simultaneously. Structured illumination (SI) has been introduced in DHM to improve the resolution, by which the resolution can be doubled. However, accurate phase-shifting is required to retrieve the low and high frequency information. Besides that, the aberration of imaging system makes DHM under SI cumbersome. This paper presents an algorithm based on PCA for DHM under SI. The aberration terms can be extracted from the first principal component of the exponential term of filtered hologram. Moreover, the low and high frequency information can be achieved from three images without prior knowledge of phase shift values. In synthesizing process, the spectrums are precisely shifted to the correct position in the spatial-frequency domain also based on the PCA. This paper verifies the feasibility of the PCA algorithm the experiment. It is an attractive and promising technology for DHM under SI.

Though widely used in image classification, convolutional neural networks (CNNs) are prone to noise interruptions, i.e. the CNN output can be drastically changed by small image noise. To improve the noise robustness, we try to integrate CNNs with wavelet by replacing the common down-sampling (max-pooling, strided-convolution, and average pooling) with discrete wavelet transform (DWT). We firstly propose general DWT and inverse DWT (IDWT) layers applicable to various orthogonal and biorthogonal discrete wavelets like Haar, Daubechies, and Cohen, etc., and then design wavelet integrated CNNs (WaveCNets) by integrating DWT into the commonly used CNNs (VGG, ResNets, and DenseNet). During the down-sampling, WaveCNets apply DWT to decompose the feature maps into the low-frequency and high-frequency components. Containing the main information including the basic object structures, the low-frequency component is transmitted into the following layers to generate robust high-level features. The high-frequency components are dropped to remove most of the data noises. The experimental results show that WaveCNets achieve higher accuracy on ImageNet than various vanilla CNNs. We have also tested the performance of WaveCNets on the noisy version of ImageNet, ImageNet-C and six adversarial attacks, the results suggest that the proposed DWT/IDWT layers could provide better noise-robustness and adversarial robustness. When applying WaveCNets as backbones, the performance of object detectors (i.e., faster R-CNN and RetinaNet) on COCO detection dataset are consistently improved. We believe that suppression of aliasing effect, i.e. separation of low frequency and high frequency information, is the main advantages of our approach. The code of our DWT/IDWT layer and different WaveCNets are available at https://github.com/CVI-SZU/WaveCNet.

Thin clouds will attenuate surface radiation, causing color distortion in remote sensing images and loss of detailed information. Aiming at the color cast phenomenon of the existing thin cloud removal methods, this paper proposes a remote sensing image thin cloud removal method based on generative adversarial networks with color consistency constraints. The RICE1 thin cloud dataset was used to verify the effectiveness of the method. CIEDE2000 was added to the evaluation index to measure the consistency of the thin cloud removal results and the color of the cloud-free image, and the reasonable value of the color constraint parameter <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\lambda$</tex> was explored. The experimental results show that the method in this paper is significantly better than homomorphic filtering, dark channel prior method and non-color-constrained generative adversarial network cloud removal method.

Summary High-resolution seismic data is the basis of fine comprehensive seismic interpretation, which lead to the key position of high-resolution seismic data processing. This paper proposes a novel method aimed at extracting the high and low frequency information by a simple and effective way. Different from the traditional spectral modeling deconvolution, this method fuses the information of different frequencies extracted from multi-order calculus to broaden frequency band. Meanwhile the constraint of desired wavelet makes the result more accurate and meaningful. Testing on practical data demonstrates it is effective and feasible.

In this paper, learning-based algorithms for image restoration and blind image restoration are proposed. Such algorithms deviate from the traditional approaches in this area, by utilizing priors that are learned from similar images. Original images and their degraded versions by the known degradation operator (restoration problem) are utilized for designing the VQ codebooks. The codevectors are designed using the blurred images. For each such vector, the high frequency information obtained from the original images is also available. During restoration, the high frequency information of a given degraded image is estimated from its low frequency information based on the codebooks. For the blind restoration problem, a number of codebooks are designed corresponding to various versions of the blurring function. Given a noisy and blurred image, one of the codebooks is chosen based on a similarity measure, therefore providing the identification of the blur. To make the restoration process computationally efficient, the principal component analysis (PCA) and VQ-nearest neighbor approaches are utilized. Simulation results are presented to demonstrate the effectiveness of the proposed algorithms.

Single image dehazing is a challenging task due to the severe degradations caused by the particles in the air. Recently, various CNN-based methods have been proposed and they have achieved promising results on some dehazing tasks. However, the existing end-to-end dehazing networks process high-frequency information and low-frequency information at the same time. Therefore, most dehazing methods cannot restore dehazed image with satisfying high-frequency details. In this paper, we propose a Frequency-aware deep Dual-path Feature enhancement Network (FDF-Net) to better restore the high-frequency information while removing the haze. To achieve this, we introduce a Dual-path Feature Enhancement (DFE) block, which contains two branches: one path is to remedy the missing spatial information from high-resolution features, and the other one is to obtain new features to increase the variety of features. We believe the dual-path architecture can help the first path to focus on the recovering the high-frequency information. Furthermore, to reserve more detailed image information from the features with larger resolution, we adopt a wavelet transform module during the downsampling process of the encoder module to directly pass the high frequency information to the next level. The extensive experiments show the superiority of the proposed model over previous methods on the benchmark datasets as well as real-world hazy images.

Individuals with body dysmorphic disorder (BDD) may have perceptual distortions for their appearance. Previous studies suggest imbalances in detailed relative to configural/holistic visual processing when viewing faces. No study has investigated the neural correlates of processing non-symptom-related stimuli. The objective of this study was to determine whether individuals with BDD have abnormal patterns of brain activation when viewing non-face/non-body object stimuli. Fourteen medication-free participants with DSM-IV BDD and 14 healthy controls participated. We performed functional magnetic resonance imaging (fMRI) while participants matched photographs of houses that were unaltered, contained only high spatial frequency (HSF, high detail) information or only low spatial frequency (LSF, low detail) information. The primary outcome was group differences in blood oxygen level-dependent (BOLD) signal changes. The BDD group showed lower activity in the parahippocampal gyrus, lingual gyrus and precuneus for LSF images. There were greater activations in medial prefrontal regions for HSF images, although no significant differences when compared to a low-level baseline. Greater symptom severity was associated with lower activity in the dorsal occipital cortex and ventrolateral prefrontal cortex for normal spatial frequency (NSF) and HSF images. Individuals with BDD have abnormal brain activation patterns when viewing objects. Hypoactivity in visual association areas for configural and holistic (low detail) elements and abnormal allocation of prefrontal systems for details are consistent with a model of imbalances in global versus local processing. This may occur not only for appearance but also for general stimuli unrelated to their symptoms.

Face perception in adults depends on skilled processing of interattribute distances ('configural' processing), which is disrupted for faces presented in inverted orientation (face inversion effect or FIE). Children are not proficient in configural processing, and this might relate to an underlying immaturity to use facial information in low spatial frequency (SF) ranges, which capture the coarse information needed for configural processing. We hypothesized that during adolescence a shift from use of high to low SF information takes place. Therefore, we studied the influence of SF content on neural face processing in groups of children (9-10years), adolescents (14-15years) and young adults (21-29years) by measuring event-related potentials (ERPs) to upright and inverted faces which varied in SF content. Results revealed that children show a neural FIE in early processing stages (i.e. P1; generated in early visual areas), suggesting a superficial, global facial analysis. In contrast, ERPs of adults revealed an FIE at later processing stages (i.e. N170; generated in face-selective, higher visual areas). Interestingly, adolescents showed FIEs in both processing stages, suggesting a hybrid developmental stage. Furthermore, adolescents and adults showed FIEs for stimuli containing low SF information, whereas such effects were driven by both low and high SF information in children. These results indicate that face processing has a protracted maturational course into adolescence, and is dependent on changes in SF processing. During adolescence, sensitivity to configural cues is developed, which aids the fast and holistic processing that is so special for faces.

Multispectral image fusion based on fractal features