Image Smoothing Research Articles

An Accuracy Reconfigurable Vector Accelerator based on Approximate Logarithmic Multipliers for Energy-Efficient Computing

The approximate logarithmic multiplier proposed by Mitchell provides an efficient alternative for processing dense multiplication or multiply-accumulate operations in applications such as image processing and real-time robotics. It offers the advantages of small area, high energy efficiency and is suitable for applications that do not necessarily achieve high accuracy. However, its maximum error of 11.1% makes it challenging to deploy in applications requiring relatively high accuracy. This paper proposes a novel operand decomposition method (OD) that decomposes one multiplication into the sum of multiple approximate logarithmic multiplications to widely reduce Mitchell multiplier errors while taking full advantage of its area savings. Based on the proposed OD method, this paper also proposes an accuracy reconfigurable multiply-accumulate (MAC) unit that provides multiple reconfigurable accuracies with high parallelism. Compared to a MAC unit consisting of accurate multipliers, the area is significantly reduced to less than half, improving the hardware parallelism while satisfying the required accuracy for various scenarios. The experimental results show the excellent applicability of our proposed MAC unit in image smoothing and robot localization and mapping application. We have also designed a prototype processor that integrates the minimum functionality of this MAC unit as a vector accelerator and have implemented a software-level accuracy reconfiguration in the form of an instruction set extension. We experimentally confirmed the correct operation of the proposed vector accelerator, which provides the different degrees of accuracy and parallelism at the software level.

함정의 평판형 방향타 캐비테이션 침식에 대한 모형 시험 연구

In the present study, a method of performing cavitation erosion test directly on the anodized surface of the rudder model is proposed, not applying ink or paint on its surface. An image processing technique is newly developed to quantitatively evaluate the erosion damages on the rudder model surface after erosion test. The preprocessing saturation image, image smoothing, adaptive hysteresis thresholding and eroded area detection algorithms are in the image processing program. The rudder cavitation erosion tests are conducted in the rudder deflection angle range of 0SUPo/SUP to -4SUPo/SUP, which is used to maintain a straight course at the highest speed of the targeted navy ship. In the case of the conventional flat-type full-spade rudder currently being used in the target ship, surface erosion can occur on the model rudder surface in the above rudder deflection angle range. The bubble type of cavitation occurs on rudder surface, which is estimated to be the main reason of erosion damage on the rudder surface.

LUNG CANCER DETECTION USING MORPHOLOGICAL WATERSHED OPERATIONS

Cancer is one of the most serious and widespread disease across the world accounting for a large number of deaths every year. Lung cancer is the leading cause of cancer deaths in the world, having high mortality rate. The main problem in lung cancer is that most of these cases are diagnosed at the later stages of cancer, making treatments more problematic and reducing the survival chances. The survival rate of Cancer patients’ increment from 14-49% if the illness is recognized in time. Early discovery of lung cancer can expand the possibility of survival among individuals. A CT scan is more likely to show lung tumors than routine chest x-rays as it can show the size, shape, and position of any lung tumor and can help find enlarged lymph nodes that might contain cancer that has spread. Visual interpretation of CT scan can be difficult and error prone. Hence image processing techniques are used to detect the cancer nodules at an early stage. This paper deals with an automated approach for early detection of lung cancer. This algorithm is proposed using methods such as Median Filtering for Image Smoothing, Contrast Adjustment for enhancing the image, Segmentation using Morphological Watershed Operations and Otsu’s Thresholding. Key Words: CT, Median Filtering, Contrast Adjustment, Morphological Watershed Operations, Otsu Thresholding.

Evaluating increases in sensitivity from NORDIC for diverse fMRI acquisition strategies

As the neuroimaging field moves towards detecting smaller effects at higher spatial resolutions, and faster sampling rates, there is increased attention given to the deleterious contribution of unstructured, thermal noise. Here, we critically evaluate the performance of a recently developed reconstruction method, termed NORDIC, for suppressing thermal noise using datasets acquired with various field strengths, voxel sizes, sampling rates, and task designs.Following minimal preprocessing, statistical activation (t-values) of NORDIC processed data was compared to the results obtained with alternative denoising methods. Additionally, we examined the consistency of the estimates of task responses at the single-voxel, single run level, using a finite impulse response (FIR) model. To examine the potential impact on effective image resolution, the overall smoothness of the data processed with different methods was estimated. Finally, to determine if NORDIC alters or removes temporal information important for modeling responses, we employed an exhaustive leave-p-out cross validation approach, using FIR task responses to predict held out timeseries, quantified using R2.After NORDIC, the t-values are increased, an improvement comparable to what could be achieved by 1.5 voxels smoothing, and task events are clearly visible and have less cross-run error. These advantages are achieved with smoothness estimates increasing by less than 4%, while 1.5 voxel smoothing is associated with increases of over 140%. Cross-validated R2s based on the FIR models show that NORDIC is not measurably distorting the temporal structure of the data under this approach and is the best predictor of non-denoised time courses. The results demonstrate that analyzing 1 run of data after NORDIC produces results equivalent to using 2 to 3 original runs and that NORDIC performs equally well across a diverse array of functional imaging protocols.Significance Statement: For functional neuroimaging, the increasing availability of higher field strengths and ever higher spatiotemporal resolutions has led to concomitant increase in concerns about the deleterious effects of thermal noise. Historically this noise source was suppressed using methods that reduce spatial precision such as image blurring or averaging over a large number of trials or sessions, which necessitates large data collection efforts. Here, we critically evaluate the performance of a recently developed reconstruction method, termed NORDIC, which suppresses thermal noise. Across datasets varying in field strength, voxel sizes, sampling rates, and task designs, NORDIC produces substantial gains in data quality. Both conventional t-statistics derived from general linear models and coefficients of determination for predicting unseen data are improved. These gains match or even exceed those associated with 1 voxel Full Width Half Max image smoothing, however, even such small amounts of smoothing are associated with a 52% reduction in estimates of spatial precision, whereas the measurable difference in spatial precision is less than 4% following NORDIC.

Soft clustering based on high- and low-level features for image smoothing

Edge-preserving image smoothing is essential for computational imaging. Traditional filters are based on low-level image gradients; however, the edges defined on gradients do not align with the contours of human perception, thus suffering from certain limitations. We propose an image filter based on a soft clustering model that combines high-level semantics (derived from instance segmentation) with low-level features (intensities). The proposed filter works by first performing soft clustering on the input image according to the high- and low-level features to derive the affinity matrices, which are then fused for image smoothing. Experiment results indicate the advantages of the proposed filter in a variety of applications, including image smoothing, flash/non-flash fusion, detail enhancing, image dehazing, and depth upsampling. Furthermore, the proposed filter is efficient, and it takes 1.75 s to process a color image with 1 megapixel on a modern desktop.

Segmenting lung parenchyma from CT images with gray correlation‐based clustering

AbstractLung segmentation, a prerequisite step of lung disease detection in computer‐aided diagnosis system, is a challenging task because of noises, complex structures, as well as large individual differences of lung CT scans. Here, an automatic algorithm for segmenting lungs from thoracic CT images accurately is presented. This scheme consists of three principal steps: image preprocessing, lung extracting and contour correcting. To cope with inhomogeneous intensities of CT images, a novel preprocessing approach based on empirical mode decomposition and bilateral filter is proposed, which has abilities of denoising, smoothing and edge keeping. Lung region is then extracted with a novel gray correlation‐based clustering approach. A new lung contour correction technology is finally employed to repair the concave regions caused by pulmonary nodules, vessels and so on. Experimental results show that the preprocessing approach outperforms other methods on image denoising and smoothing. Meanwhile, the lung segmentation algorithm is tested on a group of lung CT images affected with interstitial lung diseases and achieves a high segmentation accuracy. Compared with several existing lung segmentation methods, this algorithm exhibits a better performance on lung segmentation.

A Framework for Using UAVs to Detect Pavement Damage Based on Optimal Path Planning and Image Splicing

In order to investigate the use of unmanned aerial vehicles (UAVs) for future application in road damage detection and to provide a theoretical and technical basis for UAV road damage detection, this paper determined the recommended flight and camera parameters based on the needs of continuous road image capture and pavement disease recognition. Furthermore, to realize automatic route planning and control, continuous photography control, and image stitching and smoothing tasks, a UAV control framework for road damage detection, based on the Dijkstra algorithm, the speeded-up robust features (SURF) algorithm, the random sampling consistency (RANSAC) algorithm, and the gradual in and out weight fusion method, was also proposed in this paper. With the Canny operator, it was verified that the road stitched long image obtained by the UAV control method proposed in this paper is applicable to machine learning pavement disease identification.

Impact of γ factor in the penalty function of Bayesian penalized likelihood reconstruction (Q.Clear) to achieve high-resolution PET images

BackgroundThe Bayesian penalized likelihood PET reconstruction (BPL) algorithm, Q.Clear (GE Healthcare), has recently been clinically applied to clinical image reconstruction. The BPL includes a relative difference penalty (RDP) as a penalty function. The β value that controls the behavior of RDP determines the global strength of noise suppression, whereas the γ factor in RDP controls the degree of edge preservation. The present study aimed to assess the effects of various γ factors in RDP on the ability to detect sub-centimeter lesions.MethodsAll PET data were acquired for 10 min using a Discovery MI PET/CT system (GE Healthcare). We used a NEMA IEC body phantom containing spheres with inner diameters of 10, 13, 17, 22, 28 and 37 mm and 4.0, 5.0, 6.2, 7.9, 10 and 13 mm. The target-to-background ratio of the phantom was 4:1, and the background activity concentration was 5.3 kBq/mL. We also evaluated cold spheres containing only non-radioactive water with the same background activity concentration. All images were reconstructed using BPL + time of flight (TOF). The ranges of β values and γ factors in BPL were 50–600 and 2–20, respectively. We reconstructed PET images using the Duetto toolbox for MATLAB software. We calculated the % hot contrast recovery coefficient (CRChot) of each hot sphere, the cold CRC (CRCcold) of each cold sphere, the background variability (BV) and residual lung error (LE). We measured the full width at half maximum (FWHM) of the micro hollow hot spheres ≤ 13 mm to assess spatial resolution on the reconstructed PET images.ResultsThe CRChot and CRCcold for different β values and γ factors depended on the size of the small spheres. The CRChot, CRCcold and BV increased along with the γ factor. A 6.2-mm hot sphere was obvious in BPL as lower β values and higher γ factors, whereas γ factors ≥ 10 resulted in images with increased background noise. The FWHM became smaller when the γ factor increased.ConclusionHigh and low γ factors, respectively, preserved the edges of reconstructed PET images and promoted image smoothing. The BPL with a γ factor above the default value in Q.Clear (γ factor = 2) generated high-resolution PET images, although image noise slightly diverged. Optimizing the β value and the γ factor in BPL enabled the detection of lesions ≤ 6.2 mm.

(Retracted) Simulating human visual perception behaviors for video quality assessment

Multimedia video quality evaluation system plays an important role in modern Internet applications, such as WebRTC-based video conference. However, video quality may be affected during transmission because of noise or network congestion. Constructing an end-to-end intelligent video transmission monitoring system is of great significance. We design a system based on no-reference quality assessment, which can be used to evaluate the video-based object quality evaluation. More specifically, we first leverage Gabor filter for video segmentation. Then we conduct image smoothness estimation to measure the correlation between two adjacent frames. Subsequently, we consider two states of video distortion: compression and video transmission. We model these two states, respectively. We compare our VQA algorithm with a set of well-known NR-IQA. The reported VQA accuracies have clearly shown the superiority of our proposed video quality predictor.

Descreening of dispersed-dot halftone images with generative adversarial network

Halftoning a continuous-tone image inherently results in loss of information, which makes the inverse process, descreening, a challenging problem. Current state-of-the-art descreening algorithms have two issues: first, they mostly are PSNR-oriented reconstruction algorithms, which tend to generate piecewise smooth images that do not appear realistic due to their lack of texture. Furthermore, these algorithms are typically trained with halftone images generated from the Floyd-Steinberg error diffusion algorithm, which is not an optimal choice since the algorithm is known to generate visible artifacts in the halftone image. We address these issues by the following: first, we propose a new descreening algorithm based on conditional generative adversarial networks (cGAN) that generate descreened images with abundant texture resulting in more realistic appearance. Next, we propose using the direct binary search (DBS) algorithm instead of Floyd-Steinberg error diffusion for generating the halftone images, since it is known to generate halftone images without visible artifacts. Both qualitative and quantitative comparisons show that our algorithm outperforms state-of-the-art descreening algorithms significantly.

Edge preserving image denoising with deep networks and ensemble convolutional neural network architectures

In the field of visual perception, the edges of images tend to be rich in effective visual stimuli, which contribute to the neural network's understanding of various scenes. Image smoothing is an image processing method used to highlight the wide area, low-frequency components, and main part of the image or to suppress image noise and high-frequency interference components, which can make the image's brightness smooth and gradual, reduce the abrupt gradient, and improve the image quality. Reducing noise is treated as one of the important problems in image processing. At the same time, preserving the edges of objects is of critical importance to protect the visual appearance of the objects. Deep networks have marked a trend in computer vision applications and this paper presents a customized Gaussian noise minimizing network with edge preserving filers. Here, ensemble architecture of convolutional neural network is used in minimizing the Gaussian noise and image denoising. The ensemble architecture is combined with VGG-19 and Xception design of CNN. The ensemble Convolutional Neural Networks (CNNs) are classified for Gaussian noisy images, real noisy images, blind denoising, and hybrid noisy images, representing the combination of noisy, blurred, and lowresolution images. Following the classification, motivations and principles of various deep learning methods are analyzed. Subsequently, a comparison of state-of-the-art methods on public denoising datasets is conducted, considering both quantitative and qualitative analyses. The experimental analysis is carried out in terms of PSNR, accuracy, precision, recall and F-measure.

A Hybrid Regularization-Based Multi-Frame Super-Resolution Using Bayesian Framework

The prime purpose for the image reconstruction of a multi-frame super-resolution is to reconstruct a higher-resolution image through incorporating the knowledge obtained from a series of relevant low-resolution images, which is useful in numerous fields. Nevertheless, super-resolution image reconstruction methods are usually damaged by undesirable restorative artifacts, which include blurring distortion, noises, and stair-casing effects. Consequently, it is always challenging to achieve balancing between image smoothness and preservation of the edges inside the image. In this research work, we seek to increase the effectiveness of multi-frame super-resolution image reconstruction by increasing the visual information and improving the automated machine perception, which improves human analysis and interpretation processes. Accordingly, we propose a new approach to the image reconstruction of multi-frame super-resolution, so that it is created through the use of the regularization framework. In the proposed approach, the bilateral edge preserving and bilateral total variation regularizations are employed to approximate a high-resolution image generated from a sequence of corresponding images with low-resolution to protect significant features of an image, including sharp image edges and texture details while preventing artifacts. The experimental results of the synthesized image demonstrate that the new proposed approach has improved efficacy both visually and numerically more than other approaches.

Image enhancement based on skin-colour segmentation and smoothness

The image restoration tasks represented by image denoising, super-resolution and image deblurring have a wide range of application background, and have become a research hotspot in academia and business circles. A novel image enhancement algorithm based on skin texture preservation is proposed in this paper. The mask has been obtained using the Gaussian fitting, which can have a box blur for many times and for skin feathering. The denoising smoothing image is fused with the original image mask to preserve the hair details of the original image and enhance the edge details of the contour, so as to provide more effective information for the extraction of edge features. Compared with different methods of image smoothing algorithms, this algorithm is more effective in smoothing the skin edge contour and achieving better detection of images. Experimental results show that the proposed algorithm has strong adaptive capacity and significant effect on most images detection. Specifically, it can moderately smooth the edges of the areas with many details, leaving no traces of an artificial process. The proposed algorithm with image enhancement has a wide range of practicality.

A novel dual-fusion algorithm of single image dehazing based on anisotropic diffusion and Gaussian filter

Dark channel prior (DCP) is a widely used method in single image dehazing technology. Here, we propose a novel dual-fusion algorithm of single image dehazing based on anisotropic diffusion and Gaussian filter to suppress the halo effect or colour distortion in traditional DCP algorithms. Anisotropic diffusion is used to edge-preserving smooth images and a Gaussian filter is to smooth the local white objects. A dual-fusion strategy is conducted to optimise the atmospheric veil. Besides, the fast explicit diffusion (FED) scheme is used to accelerate the numerical solution of the anisotropic diffusion to reduce time consumption. The subjective and objective evaluation of the experiment shows that the proposed algorithm can effectively suppress the halo effect and colour distortion, and has good dehazing performance on evaluation metrics. The proposed algorithm also reduces the time consumption by 54.2% than DCP with guided filter. This study provides an effective solution for single image dehazing.

Parameter-free restoration of piecewise smooth images

Parameter-free restoration of piecewise smooth images

Focus on lasers, imaging, microscopy, and nanoscience

Focus on lasers, imaging, microscopy, and nanoscience

Bernstein inequality for multivariate functions with smooth Fourier images

UDC 517.5 Let K be a compact set in \BbbR n with ( O ) -property and let 1 ≤ p ≤ ∞ . Then there exists a constant C K < ∞ independent of f and α such that ‖ D α f ‖ p ≤ C K sup ξ ∈ K | ξ α | ‖ f ‖ ℋ p , K ,3 for all α ∈ ℤ + n and f ∈ ℋ p , K ,3 , where ℋ p , K ,3 = { f ∈ L p ( \BbbR n ) : s u p p f ^ ⊂ K , D ( 3,3 , … ,3 ) f ^ ∈ C ( \BbbR n ) } , , and f ^ is the Fourier transform of f . Note that K is said to have the ( O ) -property if there exists a constant C > 0 such that for all α ∈ ℤ + n and j = 1,2 , … , n .

Hyperparameter selection for Discrete Mumford–Shah

This work focuses on a parameter-free joint piecewise smooth image denoising and contour detection. Formulated as the minimization of a discrete Mumford-Shah functional and estimated via a theoretically grounded alternating minimization scheme, the bottleneck of such a variational approach lies in the need to fine-tune their hyperparameters, while not having access to ground truth data. To that aim, a Stein-like strategy providing optimal hyperparameters is designed, based on the minimization of an unbiased estimate of the quadratic risk. Efficient and automated minimization of the estimate of the risk crucially relies on an unbiased estimate of the gradient of the risk with respect to hyperparameters. Its practical implementation is performed using a forward differentiation of the alternating scheme minimizing the Mumford-Shah functional, requiring exact differentiation of the proximity operators involved. Intensive numerical experiments are performed on synthetic images with different geometry and noise levels, assessing the accuracy and the robustness of the proposed procedure. The resulting parameter-free piecewise-smooth estimation and contour detection procedure, not requiring prior image processing expertise nor annotated data, can then be applied to real-world images.

Easy2Hard: Learning to Solve the Intractables From a Synthetic Dataset for Structure-Preserving Image Smoothing.

Image smoothing is a prerequisite for many computer vision and graphics applications. In this article, we raise an intriguing question whether a dataset that semantically describes meaningful structures and unimportant details can facilitate a deep learning model to smooth complex natural images. To answer it, we generate ground-truth labels from easy samples by candidate generation and a screening test and synthesize hard samples in structure-preserving smoothing by blending intricate and multifarious details with the labels. To take full advantage of this dataset, we present a joint edge detection and structure-preserving image smoothing neural network (JESS-Net). Moreover, we propose the distinctive total variation loss as prior knowledge to narrow the gap between synthetic and real data. Experiments on different datasets and real images show clear improvements of our method over the state of the arts in terms of both the image cleanness and structure-preserving ability. Code and dataset are available at https://github.com/YidFeng/Easy2Hard.

Evaluating the sensitivity of Centiloid quantification to pipeline design and image resoloution

AbstractBackgroundWider use of amyloid PET in clinical routine requires validated and standardized quantification methods. Therefore, the impact of pipeline design and post‐processing smoothing on Centiloid(CL) quantification was evaluated.MethodThis study has been performed using 725 amyloid PET scans from the two AMYPAD trials(PNHS and DPMS) that were acquired with either [18F]flutemetamol or [18F]florbetaben(Table1). Using SPM12, 32 pipelines were created and calibrated to render CL with combinations of: 4 reference regions (Whole Cerebellum, Cerebellar Gray, Whole Cerebellums+Brainstem, Pons) 2 target VOI (standard GAAIN vs subject‐based) 2 reference region type (standard GAAIN vs subject‐based) 2 analysis spaces (native vs MNI) The impact of the pipeline design factors on CL values assessed using Generalized Estimating Equation(GEE) analyses. Individual morphometric variables including total intracranial volume(TIV), and grey/white matter volumes in the reference and target regions implemented in the model. To mimic differences in image resolution, PET scans were smoothed by 3mm and 6mm Gaussian filters. CL was calculated for original and corresponding smoothed PET scans. The agreement between CL values with/without smoothing evaluated using the intra‐class correlation coefficient(ICC). Bland‐Altman plots were used to demonstrate mean absolute differences between CL values after smoothing. Differences above 5CL were considered relevant.ResultThere was no statistically significant relationship for CL as a function of tracer, analysis space or target region, while RR selection and its definition did have a significant impact on CL values(Table2). More specifically, the pons provided the lowest mean CL(25.06±4.51) compared to the whole‐cerebellum(35.04±4.25) and the cerebellum grey‐matter(36.99±4.16)(Table3). Using GAAIN RR provided slightly higher mean CL values (ΔCL 4.15±1.04) compared to a subject‐based method (Table4). ICC between CL values of non‐smoothed vs smoothed PET images was 0.99 for both post‐smoothing filters(p<0.001). Differences in image resolution had minimal effect on CL values for individuals with low to moderate levels of amyloid while the effect became higher for individuals with the higher amyloid burden(Figure1)ConclusionCL quantification is robust against differences in individual morphometric measures, analysis space, image resolution, tracer and target VOI. However, reference region choice can significantly influence CL values, with the pons showing the highest discrepancies relative to other reference regions.