High-contrast Edges Research Articles

Vehicle detection in varied weather conditions using enhanced deep YOLO with complex wavelet

Traffic congestion is prevalent in many major and medium-sized cities throughout different countries in contemporary society. In traffic images, various multi-sized vehicles are tightly clustered together and obstructed from one another. Identifying vehicles in such instances is crucial for urban traffic surveillance, safety monitoring, and legal concerns but it also presents major challenges. The remarkable detection accuracy and efficiency of deep learning-based systems have led to their recent and extensive use in vehicle identification. There are significant advanced YOLO models with different backbone architectures and frameworks developed for vehicle detection. Yet, the performance of YOLO variants are facing the challenges of handling false detection against occluded and densely sophisticated scenarios. The proposed model is developed to address such types of limitations, for example; dynamic illumination, noisy images, and scale sensitivity to improve the vehicle detection rate in different traffic scenarios and varying weather conditions. The proposed study employs an improved YOLOv4 to identify moving vehicles in different lighting conditions including daylight, cloudy, rainy, and night. For hybridization, three techniques are utilized such as the Multiscale Retinex, Dual tree complex wavelet transform (DTCWT), and Pulse Coupled Neural Networks (PCNN). The DTCWT is employed for multiscale decomposition and to denoise the complex high frequency subband information, then the denoised subbands are reconstructed into a denoised image. The Multiscale retinex is utilized to reduce the halo artifacts on high-contrast edges and maintain the balance with dynamic range compression and color reproduction. The synchronizing pulse burst property of PCNN is used to detect the isolated noisy pixels and modify the detected noisy pixels. From the results it is worth noting that the developed model surpasses state-of-the-art methods in sunny, night, cloudy, and rainy modes. The proposed method using the DTCWT technique can detect the vehicles with mAP of 91.09% and 35FPS.

Infrared and visible image fusion in a rolling guided filtering framework based on deep feature extraction

AbstractTo preserve rich detail information and high contrast, a novel image fusion algorithm is proposed based on rolling-guided filtering combined with deep feature extraction. Firstly, input images are filtered to acquire various scales decomposed images using rolling guided filtering. Subsequently, PCANet is introduced to extract weight maps to guide base layer fusion. For the others layer, saliency maps of input images are extracted by a saliency measure. Then, the saliency maps are optimized by guided filtering to guide the detail layer fusion. Finally, the final fusion result are reconstructed by all fusion layers. The experimental fusion results demonstrate that fusion algorithm in this study obtains following advantages of rich detail information, high contrast, and complete edge information preservation in the subjective evaluation and better results in the objective evaluation index. In particular, the proposed method is 16.9% ahead of the best comparison result in the SD objective evaluation index.

The Fusion of Infrared and Visible Images via Feature Extraction and Subwindow Variance Filtering

This paper presents a subwindow variance filtering algorithm for fusing infrared and visible light images, with the goal of addressing challenges related to blurred details, low contrast, and missing edge features. First, images to be fused are subjected to multilevel decomposition using a subwindow variance filter, resulting in corresponding base and multiple detail layers. PCANet extracts features from the base layer and obtains corresponding weight maps that guide the fusion process. A saliency measurement method is proposed for detail‐level fusion to extract saliency maps from the source image. The saliency maps should be compared in order to obtain the initial weight map, which is then optimized using guided filtering technology to guide the fusion of detail layers. Finally, the information of the base layer and the detail layer after fusion is superimposed to obtain an ideal fusion result. The proposed algorithm is evaluated through subjective and objective measures, including information entropy, mutual information, multiscale structural similarity measurement, standard deviation, and visual information fidelity. The results demonstrate that the proposed algorithm achieves rich detail information, high contrast, and good edge information retention, making it a promising approach for infrared and visible image fusion.

Fourier phase analysis combined with a fusion scheme in long pulse thermography

Fourier phase analysis (FPA) is a common routine in long pulse thermography (LPT). However, the change of appearance in maximum phase difference in the Fourier domain caused by different depths of defects results in low contrast images with fuzzy defect edges, limiting its application majorly. In this work, the detection capability of FPA is studied by modelling the image sequence length as a function of the defect depth, thereby quantifying the expected image sequence length for inspecting the defects at a specific depth and then, a strategy for selecting the optimal image sequence length is proposed by establishing the relationship between the maximum phase difference and the image sequence length. Followed by an image fusion procedure, the phase difference map with high-contrast defect edges can be obtained from the optimal image sequence. In addition, we show that the precooling treatment before detection is an effective way to improve the image quality for LPT. Experiments have been conducted in identifying defects in a typical glass fiber reinforced polymer (GFPR) with size of 200 × 150 × 6 mm3, where continuous thermal excitation for 100 s was provided by a pair of 1 kW halogen lamps and the images were recorded by an infrared camera with wavelength band of 8–14 μm. The resulting phase difference maps clearly present the edges of all defects with a signal-to-noise ratio greater than 43 dB, demonstrating the effectiveness of the proposed method in improving the image quality in LPT.

Brightness segmentation-based plateau histogram equalization algorithm for displaying high dynamic range infrared images

Third-generation thermal cameras produce images with high dynamic range (HDR), low contrast, and blurry edges, which makes them difficult to visualize on traditional display devices. Thus, tone mapping methods are required to adapt the recorded signal to the display in order to maintain, and possibly improve, object’s visibility and contrast. Since the traditional global tone mapping methods cannot take account of the trade-off between clearly displaying the details of dark regions and bright regions of the infrared image, this paper proposes a global tone mapping algorithm based on brightness segmentation for the enhancement and display of HDR infrared images. Firstly, the HDR infrared image is divided into different brightness regions in accordance with the characteristics of the human visual system’s ability to perceive brightness. Secondly, the determination of the boundary value of each brightness region is optimized by combining the gray statistics of the HDR infrared image, to achieve the adaptive brightness region segmentation of the infrared image from various scenes. Thirdly, the plateau histogram equalization method is designed to enhance different brightness regions separately, according to their brightness levels. Finally, the enhanced brightness regions are recombined to produce a more visually pleasing output image. The effectiveness of the proposed method is analyzed of HDR infrared images taken from different scenes. A comparison of the proposed method with well-established global tone mapping techniques is provided, along with subjective and objective evaluation methods that demonstrate the potential benefits of this method in enhancing brightness, compressing dynamic range, and improving global contrast.

Amp-vortex edge-camera: a lensless multi-modality imaging system with edge enhancement.

We demonstrate a lensless imaging system with edge-enhanced imaging constructed with a Fresnel zone aperture (FZA) mask placed 3 mm away from a CMOS sensor. We propose vortex back-propagation (vortex-BP) and amplitude vortex-BP algorithms for the FZA-based lensless imaging system to remove the noise and achieve the fast reconstruction of high contrast edge enhancement. Directionally controlled anisotropic edge enhancement can be achieved with our proposed superimposed vortex-BP algorithm. With different reconstruction algorithms, the proposed amp-vortex edge-camera in this paper can achieve 2D bright filed imaging, isotropic, and directional controllable anisotropic edge-enhanced imaging with incoherent light illumination, by a single-shot captured hologram. The effect of edge detection is the same as optical edge detection, which is the re-distribution of light energy. Noise-free in-focus edge detection can be achieved by using back-propagation, without a de-noise algorithm, which is an advantage over other lensless imaging technologies. This is expected to be widely used in autonomous driving, artificial intelligence recognition in consumer electronics, etc.

Pixel-wise structured light calibration method with a color calibration target.

We propose to use a calibration target with a narrow spectral color range for the background (e.g., from blue) and broader spectral color range for the feature points (e.g., blue + red circles), and fringe patterns matching the background color for accurate phase extraction. Since the captured fringe patterns are not affected by the high contrast of the calibration target, phase information can be accurately extracted without edging artifacts. Those feature points can be clearly "seen" by the camera if the ambient light matches the feature color or without the background color. We extract each calibration pose for three-dimensional coordinate determination for each pixel, and then establish pixel-wise relationship between each coordinate and phase. Comparing with our previously published method, this method significantly fundamentally simplifies and improves the algorithm by eliminating the computational framework estimate smooth phase near high-contrast feature edges. Experimental results demonstrated the success of our proposed calibration method.

Partially coherent spiral phase contrast imaging with a 3D-like effect

Spiral phase contrast (SPC) imaging provides an effective way to isotropic edge detection for biomedical applications. Here, we present a computational SPC imaging method which combines the partially coherent illumination with the symmetry-breaking spiral phase filtering. Such an operation enables the SPC imaging to generate sharp, high-contrast and pseudo three-dimensional edge images that are almost free of speckle. The superior performance is demonstrated by imaging both amplitude and phase objects, and the speckle suppression is validated by comparing the coherent and partially coherent imaging results. This computational SPC method is expected to bring potential advantages to phase contrast imaging modality.

Global impacts of edge effects on species richness

Almost all world's forest areas are predicted to be near edges by 2100, making the detection and prediction of biodiversity responses to edges an urgent need. Yet, idiosyncratic observations across studies have hampered general inferences on the main drivers of edge effects. We provide a global meta-analysis on 674 forest edge-interior comparisons of plant and animal communities, considering both global and local drivers as well as their interactions, to predict the direction and magnitude of edge effects on species richness. We found a clear latitudinal gradient in both the direction and magnitude of edge effects, with edges in temperate regions having more species than interior forests, while tropical communities generally had fewer species at edges. Unexpectedly, communities near high-contrast edges had higher richness at edges more frequently than near low-contrast edges. In addition, richness decreases at edges were weaker in regions subjected to historical disturbance than regions without historical disturbance. These patterns were modulated by distance from the edge and became clearer when comparing more distant sampling locations. Our results provide possible explanations for the variable impacts of edge effects and thereby forest fragmentation on biodiversity, showing that a combination of local and global predictors are able to partially predict variation in edge effects.

Reduction of Artifacts and Edge Preservation of Underwater Images Using Deep Convolution Neural Network

Underwater exploration has been one of the areas of active research over the last few decades. Image enhancement is difficult for computer vision-based underwater research because of the degradation of the underwater environment’s images. Several types of artifact reduction approaches are already available of value averaging filters to smooth the continuity that artifact reduction approaches appear across image boundaries. While some of these approaches can somewhat reduce these unwanted artifacts’ severity, other approaches have some limitations that can cause blurring to high-contrast edges in the image. Therefore we need novel methods to overcome the theses’ drawbacks. This research work introduced a Deep Convolution Neural Network (DCNN)-based deep learning method to overcome the drawbacks of blurring to low-contrast edges in the image. The proposed system’s performance is validated through simulation, and the simulation results are obtained using Python simulation software. The obtained simulation results demonstrate the superiority of the proposed underwater image enhancement method. The number of different coefficients is compared with the results of the algorithm of the Underwater Image Colorfulness Measure (UICM), Underwater Image Sharpness Measure (UISM), Underwater Image Contrast Measure (UIconM) and Underwater Image Quality Measure (UIQM) values. After that, it was found to be an effective comparison of visualization techniques.

Free trajectory cone beam computed tomography reconstruction method for synchronous scanning of geometric calibration phantom and imaging object

In order to suppress the geometrical artifacts caused by random jitter in ray source scanning, and to achieve flexible ray source scanning trajectory and meet the requirements of task-driven scanning imaging, a method of free trajectory cone-beam computed tomography (CBCT) reconstruction is proposed in this paper. This method proposed a geometric calibration method of two-dimensional plane. Based on this method, the geometric calibration phantom and the imaging object could be simultaneously imaged. Then, the geometric parameters could be obtained by online calibration method, and then combined with the geometric parameters, the alternating direction multiplier method (ADMM) was used for image iterative reconstruction. Experimental results showed that this method obtained high quality reconstruction image with high contrast and clear feature edge. The root mean square errors (RMSE) of the simulation results were rather small, and the structural similarity (SSIM) values were all above 0.99. The experimental results showed that it had lower image information entropy (IE) and higher contrast noise ratio (CNR). This method provides some practical value for CBCT to realize trajectory freedom and obtain high quality reconstructed image.

Terahertz subwavelength edge detection based on dispersion-induced plasmons.

Terahertz imaging has recently attracted great attention owing to the abilities of high penetration and low ionizing damages. However, the low resolution and low contrast resulting from the diffraction limit and unwanted background illumination significantly hinder the extensive usage. In this Letter, we propose and numerically demonstrate a terahertz subwavelength imaging method capable of extracting only the edges and fine features of the targets. The underlying physics is the efficient transmission of the scattering evanescent waves related to key geometric information while blocking the propagating components. By exploiting the structurally induced plasmons in a bounded metallic waveguide, the transmission channel for evanescent waves is realized by hyperbolic metamaterials through periodically stacking dielectric layers. On this basis, high-contrast edge detection with a resolution up to ${0.1}\lambda$ is demonstrated at terahertz wavelengths. The proposed terahertz imaging method may find important applications in non-destructive testing, weak scattering object detection, and high-contrast microscopy.

An Improved Target Method to Quantitatively Measure the Lateral Resolution of the Confocal Raman Microscope

The lateral resolution is one of the most important hallmarks for evaluating the imaging performance of a confocal Raman microscope (CRM). A method based on the resolution test chart for testing the lateral resolution of the CRM was proposed. The test imaging target comprised a polished silicon substrate coated with a thin metallic pattern, which provided a high contrast and negligible edge effects. By using a single-bar target instead of a conventional three-bar target, quantitative measurements of the lateral resolution of the CRM were possible. Single-bar targets with different widths were used to test a CRM with a nominal lateral resolution of approximately 1 μm. The response of the CRM to the single-bar target was studied further with a theoretical model, and the relationship between the Michelson contrast of the response function and the lateral resolution was investigated. Finally, the method used to calculate the lateral resolution was described and tested and shown to have a relative repeatability of only 5.6%, which is ideal for resolution testing. Overall, our experimental results showed excellent agreement with simulation results and proved that the single-bar target method was capable of measuring the lateral resolution of CRMs with high accuracy, efficiency and reproducibility.

Anisotropic diffusion technique to eliminate speckle noise in continuous-wave Doppler ultrasound spectrogram

Paper described a simplified speckled elimination algorithm for the reconstruction of continuous-wave (CW) Doppler ultrasound spectrogram. Standard Nonlinear Anisotropic Diffusion (SNAD) technique fails to eliminate the speckle noise, blurs the interregional area of neighbouring speckle clusters and regulates the heat flow to preserve the systolic and diastolic peaks in the spectrogram. To resolve this problem, author modelled a kernel, conductance and diffusion function to regulate the blurring of speckle clusters and sharpen the high-contrast edges over low-contrast ones. Paper analysed the performance based on the variation of conductance (C), signal-to-noise ratio (SNR), peak signal-to-noise ratio (PSNR), mean square error (MSE), root mean square error (RMSE) and structural similarity index (SSIM) of each pairs of speckled and despeckled spectrograms. Quantitative analysis reveals that the minimised version of the kernel effectively regulates the flow of heat at edges with a high SNR (8 dB) and PSNR (>20 dB) for a low-frequency range of 20-25 Hz. SSIM (>0.9) shows an efficient structural reconstruction that identifies the lower and higher local SSIM indices as dark and bright pixels on SSIM maps. Hence, the algorithm performs its best at lower peak frequencies in the Doppler spectrograms.

Fast spot-scanning proton dose calculation method with uncertainty quantification using a three-dimensional convolutional neural network

This study proposes a near-real-time spot-scanning proton dose calculation method with probabilistic uncertainty estimation using a three-dimensional convolutional neural network (3D-CNN). CT images and clinical target volume contours of 215 head and neck cancer patients were collected from a public database. 1484 and 488 plans were extracted for training and testing the 3D-CNN model, respectively. Spot beam data and single-field uniform dose (SFUD) labels were calculated for each plan using an open-source dose calculation toolkit. Variable spot data were converted into a fixed-size volume hereby called a ‘peak map’ (PM). 300 epochs of end-to-end training was implemented using sets of stopping power ratio and PM as input. Moreover, transfer learning techniques were used to adjust the trained model to SFUD doses calculated with different beam parameters and calculation algorithm using only 7.95% of training data used for the base model. Finally, accuracy of the 3D-CNN-calculated doses and model uncertainty was reviewed with several evaluation metrics. The 3D-CNN model calculates 3D proton dose distributions accurately with a mean absolute error of 0.778 cGyE. The predicted uncertainty is correlated with dose errors at high contrast edges. Averaged Sørensen-Dice similarity coefficients between binarized outputs and ground truths are mostly above 80%. Once the 3D-CNN model was well-trained, it can be efficiently fine-tuned for different proton doses by transfer learning techniques. Inference time for calculating one dose distribution is around 0.8 s for a plan using 1500 spot beams with a consumer grade GPU. A novel spot-scanning proton dose calculation method using 3D-CNN was developed. The 3D-CNN model is able to calculate 3D doses and uncertainty with any SFUD spot data and beam irradiation angles. Our proposed method should be readily extendable to other setups and plans and be useful for dose verification, image-guided proton therapy, or other applications.

Perspectives and tradeoffs of absorber materials for high NA EUV lithography

Next-generation extreme ultraviolet (EUV) systems with numerical apertures of 0.55 have the potential to provide sub-8-nm half-pitch resolution. The increased importance of stochastic effects at smaller feature sizes places further demands on scanner and mask to provide high contrast images. We use rigorous mask diffraction and imaging simulation to understand the impact of the EUV mask absorber and to identify the most appropriate optical parameters for high NA EUV imaging. Simulations of various use cases and material options indicate two main types of solutions: high extinction materials, especially for lines spaces, and low refractive index materials that can provide phase shift mask solutions. EUV phase masks behave very different from phase shift masks for DUV. Carefully designed low refractive index materials and masks can open up a new path toward high contrast edge printing.

Two-dimensional optical spatial differentiation and high-contrast imaging.

Optical analog signal processing technology has been widely studied and applied in a variety of science and engineering fields, with the advantages of overcoming the low-speed and high-power consumption associated with its digital counterparts. Much attention has been given to emerging metasurface technology in the field of optical imaging and processing systems. Here, we demonstrate, for the first time, broadband two-dimensional spatial differentiation and high-contrast edge imaging based on a dielectric metasurface across the whole visible spectrum. This edge detection method works for both intensity and phase objects simply by inserting the metasurface into a commercial optical microscope. This highly efficient metasurface performing a basic optical differentiation operation opens up new opportunities in applications of fast, compactible and power-efficient ultrathin devices for data processing and biological imaging.

Noise and spatial resolution properties of a commercially available deep learning-based CT reconstruction algorithm.

To characterize the noise and spatial resolution properties of a commercially available deep learning-based computed tomography (CT) reconstruction algorithm. Two phantom experiments were performed. The first used a multisized image quality phantom (Mercury v3.0, Duke University) imaged at five radiation dose levels (CTDIvol : 0.9, 1.2, 3.6, 7.0, and 22.3mGy) with a fixed tube current technique on a commercial CT scanner (GE Revolution CT). Images were reconstructed with conventional (FBP), iterative (GE ASiR-V), and deep learning-based (GE True Fidelity) reconstruction algorithms. Noise power spectrum (NPS), high-contrast (air-polyethylene interface), and intermediate-contrast (water-polyethylene interface) task transfer functions (TTF) were measured for each dose level and phantom size and summarized in terms of average noise frequency (fav ) and frequency at which the TTF was reduced to 50% (f50% ), respectively. The second experiment used a custom phantom with low-contrast rods and lung texture sections for the assessment of low-contrast TTF and noise spatial distribution. The phantom was imaged at five dose levels (CTDIvol : 1.0, 2.1, 3.0, 6.0, and 10.0mGy) with 20 repeated scans at each dose, and images reconstructed with the same reconstruction algorithms. The local noise stationarity was assessed by generating spatial noise maps from the ensemble of repeated images and computing a noise inhomogeneity index, , following AAPM TG233 methods. All measurements were compared among the algorithms. Compared to FBP, noise magnitude was reduced on average (± one standard deviation) by 74±6% and 68±4% for ASiR-V (at "100%" setting) and True Fidelity (at "High" setting), respectively. The noise texture from ASiR-V had substantially lower noise frequency content with 55±4% lower NPS fav compared to FBP while True Fidelity had only marginally different noise frequency content with 9±5% lower NPS fav compared to FBP. Both ASiR-V and True Fidelity demonstrated locally nonstationary noise in a lung texture background at all radiation dose levels, with higher noise near high-contrast edges of vessels and lower noise in uniform regions. At the 1.0mGy dose level values were 314% and 271% higher in ASiR-V and True Fidelity compared to FBP, respectively. High-contrast spatial resolution was similar between all algorithms for all dose levels and phantom sizes (<3% difference in TTF f50% ). Compared to FBP, low-contrast spatial resolution was lower for ASiR-V and True Fidelity with a reduction of TTF f50% of up to 42% and 36%, respectively. The deep learning-based CT reconstruction demonstrated a strong noise magnitude reduction compared to FBP while maintaining similar noise texture and high-contrast spatial resolution. However, the algorithm resulted in images with a locally nonstationary noise in lung textured backgrounds and had somewhat degraded low-contrast spatial resolution similar to what has been observed in currently available iterative reconstruction techniques.

A Novel Approach for Mammogram Enhancement using Nonlinear Unsharp Masking and L0 Gradient Minimization

Abstract Mammography is an effective modality for detecting and diagnosing breast cancer. Mammogram enhancement is used to improve the local contrast of mammograms, such that lesions appear more visible in the enhanced image. This paper introduces two novel approaches for mammogram enhancement based on a new unsharp masking scheme, called nonlinear UM and L0 Gradient Minimization, NLUML0GMIN for detecting breast cancer at an early stage. Three different techniques are unified in the proposed method; i) a Nonlinear filtering process to enhance the fine details within a 3x3 local neighborhood, ii) an L0 Gradient Minimization step globally preserving the high-contrast edges while suppressing low-contrast details seen as opaque fibrous tissues; and to obtain a detail mammogram after subtracting smoothed mammogram from the original mammogram, iii) Finally, an Unsharp masking technique combining detail mammogram with filtered mammogram passed through Non-linear filter, using PLIP operators which satisfy both Weber’s law and saturation characteristics of the human visual system.. An HVS-based decomposition scheme is used to analyse and visualize the malignant areas in the enhanced mammogram. The distinct arrangement of PLIP operators in the proposed framework has given rise to two methods NLUML0GMINAUTO and NLUML0GMIN, which outperforms NLUM method and other existing techniques for mammogram enhancement. The former is good at enhancing less-dense and medium-dense mammograms whereas the later enhances unseen masses within high-dense mammograms. Experimental results have stated the significance of the proposed method in effectively diagnosing breast cancer in comparison with the previous methods for mammogram enhancement. The mini-MIAS database has categorized mammograms as fatty, fatty-glandular and dense-glandular, which has helped us in focusing on enhancing images uniquely for each category. The proposed NLUML0GMIN scheme is robust and effective in identifying cancerous masses and calcifications within dense X-ray mammograms, aiding doctors in better cancer diagnosis by saving the life of countless patients with cancer.

High-contrast anisotropic edge enhancement free of shadow effect.

We propose a Bessel-like composite vortex filter to perform high-contrast and power-controlled anisotropic edge enhancement with shadow-effect-free and low background noise. The background noise, which is commonly found and strongly decreases the filtered image quality in previous anisotropic vortex filters, is effectively reduced by suppressing the side lobes of the system point spread function, thereby increasing the image edge contrast to 0.98. The shadow effect is totally eliminated by keeping the radial symmetry of the filtering process, which makes edges sharper and improves image resolution. By introducing a weighting factor between two opposite vortex filter components, the power of edge enhancement becomes controllable. Numerical simulations and experimental results prove that the proposed filter achieves higher-contrast edge enhancement for both phase-contrast and amplitude-contrast objects.