Occlusion Removal Research Articles

Dose reference levels and clinical determinants in stroke neuroradiology interventions.

To establish dose reference levels (RLs) for stroke interventions while carefully analysing the impact of clinical and technical parameters on patient exposure. The study retrospectively analysed data from 377 stroke patients prospectively collected between 15 October 2015 and 30 March 2017 at a single, level-3 stroke centre equipped with Philips Allura Clarity systems. Local dose RLs were first derived as the 75th percentile of the dose area product (DAP), cumulative air kerma (Ka,r), fluoroscopy time (FT) and the number of images (NI). Univariate and multivariate negative binomial regressions were considered for the statistical analysis to investigate the dose variability with clinical and technical parameters such as patient's age and sex, occlusion removal technique, number of passages, single-plane or biplane equipment, etc. RESULTS: Local stroke dose RLs were derived in terms of total DAP (162 Gy cm2), Ka,r (854 mGy), FT (42 min) and NI (559). Gender (relative dose multiplier (RDM) 1.31; 95% CI 1.12-1.45), number of passages (RDM 1.22 per passage; 95% CI 1.10-1.22) and procedure success (RDM 0.52, 95% CI 0.55-0.80) proved to be key parameters affecting patient dose. Meanwhile the statistical analysis did not find any difference in relative dose received by patients owing to age, baseline NIHSS score, occlusion removal technique, posterior circulation, support of an anaesthesiologist or use of biplane equipment. Stroke dose RLs introduced in this work promote the optimisation of patient doses. Male gender, number of passages and success of recanalisation are independent key parameters affecting patient dose. • Stroke dose RLs derived in terms of total DAP (162 Gy cm 2 ), K a,r (854 mGy), FT (42 min) and NI (559) will help optimise the radiation safety of patients treated with mechanical thrombectomy. • Male gender (relative dose multiplier 1.31; 95% CI 1.12-1.45), number of passages (RDM 1.22 per passage; 95% CI 1.10-1.22) and success of recanalisation TICI score > 2b (RDM 0.52, 95% CI 0.55-0.80) are independent key parameters affecting patient dose. • Stent retriever or aspiration technique showed no significant difference in terms of the dose delivered to the patient; neither technique should be favoured for dosimetric reasons provided that there is no difference regarding clinical outcomes.

Improved SfM-Based Indoor Localization with Occlusion Removal

This article describes a novel 3D image-based indoor localization system integrated with an improved SfM (structure from motion) approach and an obstacle removal component. In contrast with existing state-of-the-art localization techniques focusing on static outdoor or indoor environments, the adverse effects, generated by moving obstacles in busy indoor spaces, are considered in this work. In particular, the problem of occlusion removal is converted into a separation problem of moving foreground and static background. A low-rank and sparse matrix decomposition approach is used to solve this problem efficiently. Moreover, a SfM with RT (re-triangulation) is adopted in order to handle the drifting problem of incremental SfM method in indoor scene reconstruction. To evaluate the performance of the system, three data sets and the corresponding query sets are established to simulate different states of the indoor environment. Quantitative experimental results demonstrate that both query registration rate and localization accuracy increase significantly after integrating the authors' improvements.

Spatial-temporal human gesture recognition under degraded conditions using three-dimensional integral imaging.

We present spatial-temporal human gesture recognition in degraded conditions including low light levels and occlusions using passive sensing three-dimensional (3D) integral imaging (InIm) system and 3D correlation filters. The 4D (lateral, longitudinal, and temporal) reconstructed data is processed using a variety of algorithms including linear and non-linear distortion-invariant filters; and compared with previously reported space-time interest points (STIP) feature detector, 3D histogram of oriented gradients (3D HOG) feature descriptor, with a standard bag-of-features support vector machine (SVM) framework, etc. The gesture recognition results with different classification algorithms are compared using a variety of performance metrics such as receiver operating characteristic (ROC) curves, area under the curve (AUC), SNR, the probability of classification errors, and confusion matrix. Integral imaging video sequences of human gestures are captured under degraded conditions such as low light illumination and in the presence of partial occlusions. A four-dimensional (4D) reconstructed video sequence is computed that provides lateral and depth information of a scene over time i.e. (x, y, z, t). The total-variation denoising algorithm is applied to the signal to further reduce noise and preserve data in the video frames. We show that the 4D signal consists of decreased scene noise, partial occlusion removal, and improved SNR due to the computational InIm and/or denoising algorithms. Finally, gesture recognition is processed with classification algorithms, such as distortion-invariant correlation filters; and STIP, 3D HOG with SVM, which are applied to the reconstructed 4D gesture signal to classify the human gesture. Experiments are conducted using a synthetic aperture InIm system in ambient light. Our experiments indicate that the proposed approach is promising in detection of human gestures in degraded conditions such as low illumination conditions with partial occlusion. To the best of our knowledge, this is the first report on spatial-temporal human gesture recognition in degraded conditions using passive sensing 4D integral imaging with nonlinear correlation filters.

Robust LSTM-Autoencoders for Face De-Occlusion in the Wild.

Face recognition techniques have been developed significantly in recent years. However, recognizing faces with partial occlusion is still challenging for existing face recognizers, which is heavily desired in real-world applications concerning surveillance and security. Although much research effort has been devoted to developing face de-occlusion methods, most of them can only work well under constrained conditions, such as all of faces are from a pre-defined closed set of subjects. In this paper, we propose a robust LSTM-Autoencoders (RLA) model to effectively restore partially occluded faces even in the wild. The RLA model consists of two LSTM components, which aims at occlusion-robust face encoding and recurrent occlusion removal respectively. The first one, named multi-scale spatial LSTM encoder, reads facial patches of various scales sequentially to output a latent representation, and occlusion-robustness is achieved owing to the fact that the influence of occlusion is only upon some of the patches. Receiving the representation learned by the encoder, the LSTM decoder with a dual channel architecture reconstructs the overall face and detects occlusion simultaneously, and by feat of LSTM, the decoder breaks down the task of face de-occlusion into restoring the occluded part step by step. Moreover, to minimize identify information loss and guarantee face recognition accuracy over recovered faces, we introduce an identity-preserving adversarial training scheme to further improve RLA. Extensive experiments on both synthetic and real data sets of faces with occlusion clearly demonstrate the effectiveness of our proposed RLA in removing different types of facial occlusion at various locations. The proposed method also provides significantly larger performance gain than other de-occlusion methods in promoting recognition performance over partially-occluded faces.

Disentangling the Modes of Variation in Unlabelled Data.

Statistical methods are of paramount importance in discovering the modes of variation in visual data. The Principal Component Analysis (PCA) is probably the most prominent method for extracting a single mode of variation in the data. However, in practice, several factors contribute to the appearance of visual objects including pose, illumination, and deformation, to mention a few. To extract these modes of variations from visual data, several supervised methods, such as the TensorFaces relying on multilinear (tensor) decomposition have been developed. The main drawbacks of such methods is that they require both labels regarding the modes of variations and the same number of samples under all modes of variations (e.g., the same face under different expressions, poses etc.). Therefore, their applicability is limited to well-organised data, usually captured in well-controlled conditions. In this paper, we propose a novel general multilinear matrix decomposition method that discovers the multilinear structure of possibly incomplete sets of visual data in unsupervised setting (i.e., without the presence of labels). We also propose extensions of the method with sparsity and low-rank constraints in order to handle noisy data, captured in unconstrained conditions. Besides that, a graph-regularised variant of the method is also developed in order to exploit available geometric or label information for some modes of variations. We demonstrate the applicability of the proposed method in several computer vision tasks, including Shape from Shading (SfS) (in the wild and with occlusion removal), expression transfer, and estimation of surface normals from images captured in the wild.

Face recognition against occlusions via colour fusion using 2D-MCF model and SRC

In this paper, we propose a new method to tackle the problem of face recognition with occlusions via effectively using colour information. As most of current existing approaches for this problem focus on occlusion detection first as a pre-processing step and then solve recognition problem via occlusion removal, we will develop an effective image representation technique using fused colour information to increase recognition accuracy without occlusion detection. For such purpose, we would first devise a novel local representation for face images because occlusion often appears locally. Technically, we use the recently proposed two dimensional multi-colour fusion (2D-MCF) model in our previous work for image representation, as such model has been proven to be able to achieve very high performance in terms of recognition accuracy. Then, we use the Partitioned sparse sensing recognition (P-SRC) approach for face recognition since this approach is robust for occlusions and does not need any information about the occluded areas as required in most of the existing works. Finally, extensive experiments conducted on four different databases show the superiority of the proposed method over several existing approaches including P-SRC, the Reconstruction based methods and other state of the art methods. The proposed method can improve the face recognition accuracy by up to 24.5%, 3.8%, 25% and 2.86% for the AR database, the Curtin database, the FRGC database and the Busphorus database, respectively.

Three-dimensional profilometric reconstruction using flexible sensing integral imaging and occlusion removal.

We present a method for three-dimensional (3D) profilometric reconstruction using flexible sensing integral imaging with object recognition and automatic occlusion removal. Two-dimensional images, known as elemental images (EIs), of a scene containing an object behind occlusion are captured by flexible sensing integral imaging using a moving camera randomly placed on a non-planar surface with unknown camera position and orientation. After 3D image acquisition, the unknown camera poses are estimated using the EIs and 3D reconstruction is performed based on flexible sensing integral imaging. Object recognition using the 3D reconstructed images is conducted to detect the object behind occlusion and estimate the object depth and position. Occlusion removal is then performed on the 2D EIs for the occluded object by computing variance maps of the scene. For each EI, occluded object pixels with low variance are replaced by object pixels from other perspectives using multi-view geometry. The new set of elemental images may be used to visualize the 3D profile of the scene containing the object without occlusion. Experiments are performed to validate the feasibility of the proposed method. To the best of our knowledge, this is the first report of applying flexible sensing integral imaging to profilometric reconstruction with object recognition and occlusion removal.

Profile preferentially partial occlusion removal for three-dimensional integral imaging.

A profile preferentially partial occlusion removal method for integral imaging is presented. The profile of the occlusion always contains details with significant texture structure, and regions with significant texture structure often lead to reliable depth estimation. Taking the advantage of the significant texture structure, the profile of occlusion is preferentially dealt with, and then the entire occlusion region is determined via regional spreading according to the accurate profile. The details of occlusion can be accurately removed and the occluded scene is also retained to the maximum degree. In our method, elemental images are integrated into a four-dimensional light field to provide consistently reliable depth estimation and occlusion decisions among all elemental images. Experimental results show that the proposed method is efficient to deal with the details of the occlusion, and it is robust for the occlusions with different kinds of texture structure.

Regeneration of elemental images in integral imaging for occluded objects using a plenoptic camera

In this Letter, we propose an elemental image regeneration method of three-dimensional (3D) integral imaging for occluded objects using a plenoptic camera. In conventional occlusion removal techniques, the information of the occlusion layers may be lost. Thus, elemental images have cracked parts, so the visual quality of the reconstructed 3D image is degraded. However, these cracked parts can be interpolated from adjacent elemental images. Therefore, in this Letter, we try to improve the visual quality of reconstructed 3D images by interpolating and regenerating virtual elemental images with adjacent elemental images after removing the occlusion layers. To prove our proposed method, we carry out optical experiments and calculate performance metrics such as the mean square error (MSE) and the peak signal-to-noise ratio (PSNR).

Neighboring Elemental Image Exemplar Based Inpainting for Computational Integral Imaging Reconstruction with Partial Occlusion

We propose a partial occlusion removal method for computational integral imaging reconstruction (CIIR) based on the usage of the exemplar based inpainting technique. The proposed method is an improved version of the original linear inpainting based CIIR (LI-CIIR), which uses the inpainting technique to fill in the data missing region. The LI-CIIR shows good results for images which contain objects with smooth surfaces. However, if the object has a textured surface, the result of the LI-CIIR deteriorates, since the linear inpainting cannot recover the textured data in the data missing region well. In this work, we utilize the exemplar based inpainting to fill in the textured data in the data missing region. We call the proposed method the neighboring elemental image exemplar based inpainting (NEI-exemplar inpainting) method, since it uses sources from neighboring elemental images to fill in the data missing region. Furthermore, we also propose an automatic occluding region extraction method based on the use of the mutual constraint using depth estimation (MC-DE) and the level set based bimodal segmentation. Experimental results show the validity of the proposed system.

Mutual constraint using partial occlusion artifact removal for computational integral imaging reconstruction

In this paper, we propose a partial occlusion removal method to obtain an artifact-free three-dimensional reconstructed image by computational integral imaging reconstruction (CIIR). The method utilizes the mutual constraints that exist between the elemental images in the elemental image array to obtain a reliable depth map. Normally, in CIIR, the elemental images are very small and have very low resolution, which makes it difficult to obtain a reliable depth field. Therefore, previous researchers proposed a subimage-based framework, where elemental images are first transformed into subimages by a subimage transform to increase the small sizes of the elemental images. After that, a stereomatching algorithm is applied on the subimages to compute the depth map from which the region of occlusion can be extracted. However, due to the inherent incorrectness of the subimages and the weak constraint used in the stereomatching algorithm, the resulting depth map includes many false regions, which result in a noisy segmentation of the occluding object region. By utilizing the mutual relationship between neighboring elemental images, a strong constraint can be applied on the computation of the depth map. This reduces the uncertainty in the computation and thus results in a reliable depth map. This is due to the fact that owing to the strong constraint, the depth map can be computed directly from the elemental images, excluding the incorrectness of the subimages. We also describe both the pixel-wise and window-based implementations of the proposed method. Experimental results verify the performance of the proposed method.

Robot Body Occlusion Removal in Omnidirectional Video Using Color and Shape Information

Robot Body Occlusion Removal in Omnidirectional Video Using Color and Shape Information

A highly accurate dense approach for homography estimation using modified differential evolution

The homography corresponds to a 3×3 matrix which transfers image points between two images of a planar scene or two images captured by cameras under purely rotational motion. Homography estimation is crucial to many computer vision tasks involving multi-view geometry. Unlike most existing algorithms using sparse information extracted from images, this paper proposes a dense homography estimation method taking into account both the color and gradient information of the whole image. In particular, homography estimation is recast as a two-objective optimization problem, which is solved within a global optimization procedure guided by sparse control points (SCPs) based on modified differential evolution (DE). In order to improve the computational efficiency, two strategies namely pre-evaluation and coarse-to-fine (C2F) search are integrated into the proposed framework. The experimental results on synthetically rendered images and real images demonstrate that the new method can consistently improve the accuracy of homography estimation compared with two most successful feature based algorithms, and the incorporated acceleration strategies are able to speed up the minimization procedure by about 15 times. The applicability of our proposed method is further demonstrated in two real-world applications, namely occlusion removal and image mosaicing.

Occlusion Removal and Filling in Disparity Map for Real Time Multiple View Synthesis

In synthesizing multiple intermediate views from two stereo using stereo-matching, we suggest an efficient method for removing occlusion. We dealt with the noise correction from stereo matching, occlusion removal for high fidelity rendering, and parallel computing for real-time synthesis. The proposed algorithm shows the better performance in rendering quality and computational speed.

Occlusion Removal Using Depth Mapping in Three-Dimensional Integral Imaging

In this paper, we present an occlusion removal method to enhance reconstructed images of occluded three-dimensional (3D) objects based on pixel depth mapping technique in 3D multi-perspective imaging. Depth mapping is achieved by minimizing the statistical variance of projection image pixels of 3D object points on different perspective images. Then depth map and variance map are utilized to classify elemental image pixels into object or occlusion classes. Based on pixel classification results, a modified reconstruction algorithm is used to reconstruct the object without the effect of occlusion. In the proposed method, the occlusion is unknown and may be arbitrarily placed in front of the scene. Experimental results are presented to evaluate the effectiveness of the proposed method.

Age‐related changes of skin blood flow during postocclusive reactive hyperemia in human

The objective was to study age-related alterations in the time-amplitude characteristics of the oscillatory components of peripheral blood flow in healthy humans during postocclusive reactive hyperemia. Forearm blood skin perfusion was studied by the method of laser Doppler flowmetry in 120 healthy volunteers, who were divided into two age groups: young (19-30 years old; n = 82) and middle-age (30-60 years old; n = 38). The forearm occlusion approach was used to reveal the reaction of the microvascular bed to transient ischemia. To estimate the age-related changes of separate oscillatory components of LDF signals during postocclusive reactive hyperemia, we applied adaptive time-amplitude wavelet analysis. A statistically significant increase in the skin blood perfusion after occlusion removal was revealed: 4-fold in the 1st and 3.5-fold in the 2nd group respectively. Both the amplitude of blood flow oscillations at rest and the postocclusive growth of the amplitude showed an age-related decline in the frequency intervals associated with the myogenic (0.052-0.145 Hz), sympathetic (0.021-0.052 Hz) and endothelial (0.0095-0.021 Hz) activity. The time-amplitude characteristics of the oscillatory components of peripheral blood flow depend on age, with the dependence becoming more evident after a transient ischemia.

FEBEMD and Exemplar based Hybrid Image Inpainting for Occlusion Removal

ABSTRACT In this paper, a hybrid method for occlusion removal using Finite element based Bi-dimensional Empirical Mode Decomposition (FE-BEMD) and Exemplar based image inpainting is discussed. Initially, the image is decomposed into Intrinsic Mode Functions (IMFs) and a Residue using FE-BEMD. Then the Exemplar based image inpainting algorithm is applied to each of the IMFs and residue and the results are added together to get the inpainted image. The results obtained shows that the proposed method works well for removing large objects as well as small damages like scratches from images. General Terms Image decomposition, occlusion removal from images. Keywords Image inpainting, Empirical Mode Decomposition, Intrinsic Mo-de Function, Delaunay Triangles, Texture synthesis, simultaneo-us structure and texture propagation. 1. INTRODUCTION Most of our historical data are in the form of images. Human and environmental factors can affect these images and damage them. The technique of filling-in the missing data in a damaged image in a visually plausible way is called image inpainting. It is also called disocclusion or image completion. Image inpainting is usually used to repair the damaged portions in a picture and also to remove unwanted objects from a picture, like overwriting. Its applications include restoration of old photographs, films and paintings and also cleaning documents. FE-BEMD is a fully data-driven, unsupervised decomposition method that decomposes images into Intrinsic Mode Functions and Residue. When all the IMFs are obtained, the data is actually decomposed into a number of frequency bands, in a nonlinear sense, as this method handles non-stationary signals. The lower IMFs pertain to the edges and higher IMFs pertain to illumination. This paper discusses a method that removes large objects as well as small damages like scratches from images, by applying exemplar based image inpainting to the IMFs and residue obtained by FE-BEMD of the images. The paper is organized as follows. Section 2 describes some of the existing inpainting methods. Section 3 deals with Finite Element based Bi-Dimensional Empirical Mode Decomposition. Section 4 describes exemplar based Image inpainting using FE-BEMD and Section 5 gives the experimental results. The last section provides the conclusion.

Predictive Factors for the Removal of "Soft" Femoropopliteal Occlusions with Intraarterial Short-term and High-Dose Urokinase Treatment.

To date, intraarterial lysis is a widely accepted form of treatment for femoropopliteal occlusions. However, the outcome of therapy has been difficult to predict. Therefore, the present study dealt with two parameters likely to influence the occlusion removal rate. These are the outflow condition distal of the occlusion, and the occlusion length. Fifty patients with subacute femoropopliteal occlusions were treated with intraarterial thrombolysis and, in case residual obstructions were left, with subsequent catheter dilation. Only "soft" occlusions that were probed by a guide wire and proved to be penetrable were chosen for treatment. The occlusion age was 3.79 +/- 4.57 weeks with 90% of cases in the range of 1 day to 6 weeks. Fibrinolytic therapy consisted of one to three one-hour series of 300,000 IU urokinase instillation directly into the occlusion material. The complete clearance rate was 34/50 = 68%. In successfully treated patients, both the pressure gradient (difference between systemic and ankle pressure) and the walking distance improved significantly. Most important factors for success or failure of treatment were the quality of calf outflow and the length of obstruction. Thus, patients with at least one calf artery open and an occlusion length of under 10 cm presented with an opening rate of 88.5%, whereas patients with no calf arteries patent and an occlusion length over 20 cm presented with an opening rate of only 20%. The age of the occlusions, all penetrable by a guide wire prior to treatment, did not influence the rate of recanalization. Intraarterial short-term and high-dose urokinase infusion is an excellent form of treatment for subacute femoropopliteal occlusions. Especially good results with a 88.5% clearance rate were achieved in patients who presented with at least one calf artery open and an occlusion length below 10 cm.

Enhanced reconstruction of partially occluded objects with occlusion removal in synthetic aperture integral imaging

Enhanced reconstruction of partially occluded objects with occlusion removal in synthetic aperture integral imaging

合成孔径集成成像中基于遮挡物去除的目标重构

合成孔径集成成像中基于遮挡物去除的目标重构