Target Image Quality Research Articles

A lightweight convolutional neural network super-resolution method to improve the quality of target image in space measurement tasks

A lightweight convolutional neural network super-resolution method to improve the quality of target image in space measurement tasks

Clear imaging method for underwater targets based on the second Lorentz depolarization index

Underwater target information is submerged in background scattered light due to the scattering effect of suspended particles in turbid water. In order to improve the quality of target imaging in turbid water, this paper proposes a method for achieving clear underwater target imaging using the second Lorentz depolarization index based on a dual division of focal plane Mueller matrix imaging system. This method can directly quantify the attenuation of light in water, and then solve a clear image of the target. Experimental results on underwater targets of different materials (plastic, paper, metal) show that the method proposed can effectively remove background scattering light while enhancing image contrast, ultimately improving the imaging quality of targets in turbid water.

Transcranial ultrafast ultrasound Doppler imaging: A phantom study

Ultrafast ultrasound Doppler imaging facilitates the assessment of cerebral hemodynamics with high spatio-temporal resolution. However, the significant acoustic impedance mismatch between the skull and soft tissue results in phase aberrations, which can compromise the quality of transcranial imaging and introduce biases in velocity and direction quantification of blood flow. This paper proposed an aberration correction method that combines deep learning-based skull sound speed modelling with ray theory to realize transcranial plane-wave imaging and ultrafast Doppler imaging. The method was validated through phantom experiments using a linear array with a center frequency of 6.25 MHz, 128 elements, and a pitch of 0.3 mm. The results demonstrated an improvement in the imaging quality of intracranial targets when using the proposed method. After aberration correction, the average locating deviation decreased from 1.40 mm to 0.27 mm in the axial direction, from 0.50 mm to 0.20 mm in the lateral direction, and the average full-width-at-half-maximum (FWHM) decreased from 1.37 mm to 0.97 mm for point scatterers. For circular inclusions, the average contrast-to-noise ratio (CNR) improved from 8.1 dB to 11.0 dB, and the average eccentricity decreased from 0.36 to 0.26. Furthermore, the proposed method was applied to transcranial ultrafast Doppler flow imaging. The results showed a significant improvement in accuracy and quality for blood Doppler flow imaging. The results in the absence of the skull were considered as the reference, and the average normalized root-mean-square errors of the axial velocity component on the five selected axial profiles were reduced from 17.67% to 8.02% after the correction.

Improving millimeter-wave imaging quality using the vortex phase method

This paper investigates a new vortex wave imaging approach to improve the imaging quality of small metal targets of size less than 1.5 mm. Antennas with different spiral phase plates are designed to efficiently transmit vortex beams with orbital angular momentums (OAMs). By analyzing the OAM spectrum of the target, it was discovered that the predominant reflection contains a particular OAM mode that carries abundant azimuthal information. This can be explained by the OAM selectivity of the target and the guidance of the vortex transmitting beam. A simple reflection vortex imaging system was designed to capture the phase information. Measurement results show that the high image contrast reaches 14.9%, which is twice as high as that of the imaging without OAM. Both of simulations and experiments demonstrate that the vortex phase imaging approach proposed in this paper can effectively improve the imaging quality at 80 GHz. This approach is suitable for other millimeter wave imaging systems and is helpful to improve the resolution in anti-terrorism security checks.

Task-based detectability in anatomical background in digital mammography, digital breast tomosynthesis and synthetic mammography

Objective. Determining the detectability of targets for the different imaging modalities in mammography in the presence of anatomical background noise is challenging. This work proposes a method to compare the image quality and detectability of targets in digital mammography (DM), digital breast tomosynthesis (DBT) and synthetic mammography. Approach. The low-frequency structured noise produced by a water phantom with acrylic spheres was used to simulate anatomical background noise for the different types of images. A method was developed to apply the non-prewhitening observer model with eye filter (NPWE) in these conditions. A homogeneous poly(methyl) methacrylate phantom with a 0.2 mm thick aluminium disc was used to calculate 2D in-plane modulation transfer function (MTF), noise power spectrum (NPS), noise equivalent quanta, and system detective quantum efficiency for 30, 50 and 70 mm thicknesses. The in-depth MTFs of DBT volumes were determined using a thin tungsten wire. The MTF, system NPS and anatomical NPS were used in the NPWE model to calculate the threshold gold thickness of the gold discs contained in the CDMAM phantom, which was taken as reference. Main results. The correspondence between the NPWE model and the CDMAM phantom (linear Pearson correlation 0.980) yielded a threshold detectability index that was used to determine the threshold diameter of spherical microcalcifications and masses. DBT imaging improved the detection of masses, which depended mostly on the reduction of anatomical background noise. Conversely, DM images yielded the best detection of microcalcification s. Significance. The method presented in this study was able to quantify image quality and object detectability for the different imaging modalities and levels of anatomical background noise.

Vision-based underwater target real-time detection for autonomous underwater vehicle subsea exploration

Autonomous underwater vehicles (AUVs) equipped with online visual inspection systems can detect underwater targets during underwater operations, which is of great significance to subsea exploration. However, the undersea scene has some instinctive challenging problems, such as poor lighting conditions, sediment burial, and marine biofouling mimicry, which makes it difficult for traditional target detection algorithms to achieve online, reliable, and accurate detection of underwater targets. To solve the above issues, this paper proposes a real-time object detection algorithm for underwater targets based on a lightweight convolutional neural network model. To improve the imaging quality of underwater images, contrast limited adaptive histogram equalization with the fused multicolor space (FCLAHE) model is designed to enhance the image quality of underwater targets. Afterwards, a spindle-shaped backbone network is designed. The inverted residual block and group convolutions are used to extract depth features to ensure the target detection accuracy on one hand and to reduce the model parameter volume on the other hand under complex scenarios. Through extensive experiments, the precision, recall, and mAP of the proposed algorithm reached 91.2%, 90.1%, and 88.3%, respectively. It is also noticeable that the proposed method has been integrated into the embedded GPU platform and deployed in the AUV system in the practical scenarios. The average computational time is 0.053s, which satisfies the requirements of real-time object detection.

Measurement Matrix Optimization Based on Target Prior Information for Radar Imaging

Radar target imaging can provide important information for target monitoring and recognition. In order to reduce the data amount and save radar resources, the target imaging methods based on compressed sensing have received extensive attention. In these methods, the measurement matrix will directly affect the target imaging quality and data down-sampling rate. However, most of the existing researches on measurement matrix optimization did not take into account the full use of target characteristic information and the internal consistency between the measurement matrix structure and the data sampling process, which makes it difficult to obtain the optimal imaging quality and down-sampling rate. To solve the above problems, this paper makes full use of the target imaging prior information, and takes the minimum dimension of the measurement matrix and the maximum similarity between the obtained image and expected image as the goal of measurement matrix optimization. On this basis, aims at the kind of signal which is composed of a group of sub-pulses with stepped frequency, a joint optimization model of measurement matrix and imaging performance based on the target imaging prior is established with the consideration of the 0-1 constraint of the measurement matrix structure caused by the physical observation process of the sparse stepped frequency signals. And then the corresponding sparse reconstruction algorithm is proposed. Thus, the optimal measurement matrix and high-resolution imaging results can be obtained, and the radar data down-sampling rate can be improved significantly. Simulation results indicate the effectiveness of the proposed method.

Joint spatial structural sparsity constraint and spectral low-rank approximation for snapshot compressive spectral imaging reconstruction

Snapshot compressive spectral imaging (SCI) is a potential technology in the field of hyperspectral imaging (HSI), where multi-frame spectral images are compressed into a single snapshot measurement and collected by a 2D detector. Existing reconstruction algorithms of SCI systems do not make full use of the redundancy of hyperspectral data, artifacts and blur degrade the quality of the reconstruction target image. In this paper, an efficient algorithm, named as spatial structural sparsity and spectral low-rank priors (SSS-SLR), is proposed based on two inherent priors of hyperspectral images. Specifically, the spatial structural sparsity and spectral low-rank priori are simultaneously integrated into the SCI image restoration process to establish a variational optimization model, which can be solved via an alternating minimization algorithm. Extensive reconstruction results of hyperspectral data cube from both synthetic and real datasets demonstrate that the proposed method significantly outperforms the canonical algorithms.

(Retracted) Segmentation method for low-quality images of coal and gangue based on Retinex and local texture features with multifractal

Segmentation is very important for the automatic recognition of coal and gangue. Aiming at the problem that it is difficult to segment coal and gangue accurately from background, the segmentation method is designed for low-quality images based on local texture features with multifractal. First, the existing problems of coal and gangue images are analyzed in detail, and the preprocessing model is established according to characteristics of their surface texture, color, and noise. Second, the dynamic range compression and edge enhancement of Retinex algorithm are used to improve the contrast and image quality of target. Then, the local multifractal is used to represent the texture features. Finally, the automatic segmentation of coal and gangue is realized by feature clustering based on the local texture features. The experimental results show that the proposed method can effectively maintain local features while improving overall brightness of the targets, and the segmentation results show that this method is effective and has better performance for coal and gangue images in target and background segmentation, and also has good performance in the recognition.

RootPainter: deep learning segmentation of biological images with corrective annotation.

Convolutional neural networks (CNNs) are a powerful tool for plant image analysis, but challenges remain in making them more accessible to researchers without a machine-learning background. We present RootPainter, an open-source graphical user interface based software tool for the rapid training of deep neural networks for use in biological image analysis. We evaluate RootPainter by training models for root length extraction from chicory (Cichorium intybus L.) roots in soil, biopore counting, and root nodule counting. We also compare dense annotations with corrective ones that are added during the training process based on the weaknesses of the current model. Five out of six times the models trained using RootPainter with corrective annotations created within 2 h produced measurements strongly correlating with manual measurements. Model accuracy had a significant correlation with annotation duration, indicating further improvements could be obtained with extended annotation. Our results show that a deep-learning model can be trained to a high accuracy for the three respective datasets of varying target objects, background, and image quality with < 2 h of annotation time. They indicate that, when using RootPainter, for many datasets it is possible to annotate, train, and complete data processing within 1 d.

The Multi-domain Union Clutter Suppression Algorithm Based on Robust Principal Component Analysis

In through-the-wall imaging, the clutter cannot be eliminated completely through traditional algorithms, and seriously affects the subsequent target detection and recognition. To solve the problem, based on robust principal component analysis theory, a joint low-rank and sparse model is established in echo and image domain respectively. The models are solved by smoothing fast alternating linearization method. Then, the target images are dealt with exponentially weighted multiply multi-domain image fusion to obtain the final image. The simulation results indicate that the algorithm has great speed and accuracy with effective improvement on imaging quality of targets.

A Method for Signal Leakage Cancellation in Multistatic Subsurface SAR Imaging System

Multi-static subsurface synthetic aperture radar (SAR) is a promising tool capable of providing 3-D high resolution images of targets buried underground. The high detection capability of the system stems from the fact that the multi-static configuration allows for viewing the targets from many directions. However, a primal limiting factor for such configuration is the dominance of the direct and ground reflection signals over the scattered target signal hindering the system sensitivity. This letter presents a simple technique to overcome such drawback. By an appropriate utilization of SAR data, each adjacent sampling points are co-processed together as a two-element array in such a way as to have a null along the direction of the direct and ground reflection signals suppressing such signals and making the target signals more discernible. The technique is described in details and the potential problems arising from array processing are discussed. Additionally, a modified version of back projection imaging algorithm suitable for the array is provided. The performance of the proposed technique is assessed by simulation data of buried metallic spheres as well as measurement data for a buried metallic pipeline. Both simulation and measurements show a significant improvement in target detection and image quality.

Blind restoration of astronomical image based on deep attention generative adversarial neural network

The imaging quality of astronomical targets observed by ground-based telescopes is affected by atmospheric turbulence and the image resolution is seriously reduced. A deep attention generative adversarial network is proposed to restore the astronomical image and to learn the end-to-end imaging law between the blurred image and the ground truth image from image dataset directly. The attention mechanism module is designed to improve the performance of the network. Based on the conventional theory of atmospheric imaging of telescopes and combining optical system parameters, a series of astronomical images are simulated to establish a dataset for training networks. The proposed method is verified by simulated test image and real astronomical image. The experimental results show that the proposed method can effectively eliminate the influence of atmospheric turbulence and improve the resolution of astronomical images. We demonstrate the possible and good prospects for future applications of deep learning to high-resolution imaging of astronomical images.

Imaging targets hidden in scattering and viscous liquid-based media by combining multiple projections and applying a non-local mean filtering algorithm

Scattering and viscous (or sticky) liquid-based media such as that in the human ear can be considered a highly turbid environment through which imaging a target is challenging. Typically, reduced target visibility and image quality together with the limitation of depth through which objects may be resolved occurs because of multiple scattering events. Consequently, several approaches described in the literature have been applied to image objects hidden within turbid liquid media. In this paper, we experimentally demonstrate an approach to enhance image quality during recovery of objects hidden in sticky liquid at different viscosity levels by the fusion of single-shot projections assisted by a non-local mean (NLM) filtering algorithm. The turbidity of the medium was evaluated via attenuation coefficient by Beer's law and the viscosity was measured by a capillary viscometer device. The medium was illuminated with a laser beam and multiple images of the object were obtained from different perspectives using a lens array. Following image processing, each sub-image (projection) was digitally cropped, extracted from the array, filtered, shifted to a common center and then summed with other sub-images to form a single average image. Different setup conditions such as: medium turbidity, target shape, illumination state, and filtering processing were tested and compered during this work. A quantitative image quality metric of visibility commonly used in image processing was adopted to evaluate the quality of the reconstructed targets. The feasibility of the proposed hybrid method is supported by the experimental results which show a reconstructed image with high visibility. This study is a part of our ongoing efforts to enable imaging and reveal objects embedded in turbid media by integrating different strategies from the disciplines of: optics, astrophysics, digital image processing, and medicine.

Chirplet signal design by FPGA

The ever-expanding growth of the electronics and communications industries present new challenges for researchers. One of these challenges is the generation of the required bandwidth signal over a specific time frame that is used in a variety of contexts, particularly radar systems. To improve the range resolution in the radar along with better SNR, it is necessary to reduce the signal bandwidth and increase the peak power. There are some restrictions for narrowband signals like power limitation, pulse shaping, and the production of unwanted harmonics. So as a solution pulse compression techniques are suggested. Pulse compression is a process that modulating the transmitted pulse to achieve a wideband signal and then at the receiver, the received signal correlates with the transmitted pulse to achieve narrowband representations of data. Chirp is the most common signal used in pulse compression. The chirp signal is produced using linear frequency modulation. In this study, we attempted to add an amplitude modulation to the chirp signal and evaluate its performance by implementation on FPGA. The outcome signal is called chirplet and simulation will show that it enhance target detection and image quality in imaging radars like SAR.

Fast Generation of Deceptive Jamming Signal Against Spaceborne SAR Based on Spatial Frequency Domain Interpolation

This article proposes a novel method for the generation of the deceptive jamming signal against spaceborne SAR, which can calculate the jammer’s frequency response (JFR) efficiently. The first advantage of this algorithm is that the real-time computational complexity is independent of the number of fake scattering units, so it is particularly suitable for generating scenes containing a large number of scatterers. Besides, the imaging quality of fake targets is also improved significantly, especially in the case of squint geometries. First, using the Taylor series expansion and approximations, we derive the recurrence relationship of the slant range between the adjacent pulse repetition intervals. Based on this relationship, the JFR at each azimuth time is mapped to a special spatial spectrum, which is the 2-D Fourier transform along the range and azimuth dimensions of the JFR matrix at the time of the first pulse arrival. Therefore, after the initial JFR is calculated, the subsequent JFRs can be quickly obtained by sinc function interpolation. Then, the validity and effective region of this algorithm are estimated by the theoretical analysis of the slant range error. The simulation results and a computational complexity analysis verify that the algorithm demonstrates superior performance in terms of imaging quality and efficiency.

Camera Image Quality Tradeoff Processing of Image Sensor Re-mosaic using Deep Neural Network

Recently, with the release of 108 mega pixel resolution image sensor, the photo quality of smartphone camera, including detail, and texture, is getting much higher. This became possible only because by utilizing the remosaic technology which re-organize color filter arrays into the Bayer patterns compatible to existing Image Signal Processor (ISP) of commodity AP. However, the optimized parameter configurations of the remosaic block require lots of efforts and long tuning period in order to secure the desired image quality level and sensor characteristics. This paper proposes a deep neural network based camera auto-tuning system for the remosaic ISP block. Firstly, considering the learning phase, big image quality database is created in the random way using reference image and tuning register. Second, the virtual ISP model has been trained in order that predicts image quality by changing sensor tuning registers. Finally, the optimization layer generates the sensor remosaic parameters in order to achieve the user’s target image quality expectation. By experiment, the proposed system has been verified to secure the image quality at the level of professionally hand-tuned photography. Especially, the remosaic artifact of false color, color desaturation and line broken artifacts are improved significantly by more than 23%, 4%, and 12%, respectively.

Order‐controlled closed‐loop focal beams and resolution comparison of primary and multiple reflections for seismic acquisition geometries

ABSTRACTFocal beam analysis has built a bridge between the acquisition parameters on the surface and the image quality of underground targets. However, as a practical matter, it is still difficult to answer how to choose a proper acquisition geometry according to the complexity of medium, especially considering the contradictory effects of multiple reflections on spatial resolution as they can be considered to be either potential signal or additional noise, depending on the envisioned imaging technology. We introduce an order‐controlled, closed‐loop focal beam method in which the migration operator and the resolution function can be analysed in the process of the closed‐loop migration with full control over the order of the surface and internal multiples considered. This method highlights the effects of primary and different‐order multiple wavefields on the imaging resolution for different acquisition geometries and various overburden strata. We apply the method to analyse the predicted resolution of seismic acquisition geometries considering multiples as either noise or signal. Results show, in the acquisition geometry design, that when the primaries cannot provide a complete spatial illumination for the subsurface target, e.g. because of the limited‐aperture acquisition geometries or the complicated overburden, we should use the closed‐loop focal beam analysis to assess the contradictory effects of multiples as both signal and noise, in which the maximum order of multiples ought to be chosen according to the core aim of the acquisition analysis. We can apply the second‐order closed‐loop focal beam analysis to quantify the effects of acquisition geometries on multiple‐wave suppression and can also perform the high‐order closed‐loop focal beam analysis to quantify the effects of acquisition geometries on high‐resolution imaging (migration). This method can also be used to choose the optimal order of multiples in the closed‐loop migration.

Morphological Band Registration of Multispectral Cameras for Water Quality Analysis with Unmanned Aerial Vehicle

Multispectral imagery contains abundant spectral information on terrestrial and oceanic targets, and retrieval of the geophysical variables of the targets is possible when the radiometric integrity of the data is secured. Multispectral cameras typically require the registration of individual band images because their lens locations for individual bands are often displaced from each other, thereby generating images of different viewing angles. Although this type of displacement can be corrected through a geometric transformation of the image coordinates, a mismatch or misregistration between the bands still remains, owing to the image acquisition timing that differs by bands. Even a short time difference is critical for the image quality of fast-moving targets, such as water surfaces, and this type of deformation cannot be compensated for with a geometric transformation between the bands. This study proposes a novel morphological band registration technique, based on the quantile matching method, for which the correspondence between the pixels of different bands is not sought by their geometric relationship, but by the radiometric distribution constructed in the vicinity of the pixel. In this study, a Micasense Rededge-M camera was operated on an unmanned aerial vehicle and multispectral images of coastal areas were acquired at various altitudes to examine the performance of the proposed method for different spatial scales. To assess the impact of the correction on a geophysical variable, the performance of the proposed method was evaluated for the chlorophyll-a concentration estimation. The results showed that the proposed method successfully removed the noisy spatial pattern caused by misregistration while maintaining the original spatial resolution for both homogeneous scenes and an episodic scene with a red tide outbreak.

Fast overlapped subaperture algorithm for high-squint spotlight SAR imaging

ABSTRACT The overlapped subaperture algorithm (OSA) is popular for high-squint spotlight synthetic aperture radar (SAR) imaging due to its ability of dealing with a much larger swath than the primary polar format algorithm. However, the traditional OSA suffers from the problem of low efficiency because of the high subaperture overlapping ratio (SOR) required to get rid of spurious targets. The contribution of this study is to loosen the original SOR limit by embedding an additional spatially variant apodization (SVA) into the architecture, resulting in a new fast OSA with lighter computational load and hence higher processing efficiency. Specifically, the algorithm is modified mainly from two aspects: First, a new spatial variant quadratic phase error (QPE) filter is induced to enable effective SVA implementation; Second, a new squint subaperture stitch method is induced to accommodate the subaperture extraction with the new data format led by the new QPE filter. The proposed approach is evaluated by both the point target simulations and the real airborne SAR data in aspects of spurious target suppression, image quality and processing efficiency.