Rough Segmentation Research Articles

Behavior-induced health condition monitoring of caged chickens using binocular vision

The behaviors of chicken can reflect its health condition. However, the existing behaviors information acquisition directly depending on human observation is tedious, laborious and inaccurate. This paper aims to present an automatic behaviors information monitoring method for caged chicken using the binocular vision system. First, a 2-Dimension image segmentation algorithm based on active contour model was improved to accurately segment the chicken head and body. Specifically, ‘S’ weight of HSV color model and ‘a’ weight of Lab color model were used as the fundamental images for chicken body and head segmentation, respectively. Binary operation and morphology processing were utilized to obtain the rough segmentation results, which were regarded as the initial contour of the active contour model. In order to eliminate the interference of cage, we improved the active contour model by integrating with gray morphology and Gaussian filter operations. Second, the 3D image reconstruction was carried out using pairs of rectified images obtained from the binocular cameras. Thus, 3D contours of chicken body and head can be extracted from the above 2D segmentation and 3D reconstruction results. Third, motion parameters derived from eating and drinking behaviors of chicken were computed from the 3D space. Finally, the relationship between behaviors information and health condition was explored. The experimental results showed that the proposed method can monitor the health condition of the caged chicken in real-time.

Sparse patch‐based representation with combined information of atlas for multi‐atlas label fusion

To obtain a higher accuracy in the multi-atlas patch-based label fusion method, it is essential to have the accurate similarity measure of selected patches. In this study, the authors propose a new sparse patch-based representation method using a local binary texture (LBT) in the atlas image and atlas label information for the multi-atlas label fusion. In the proposed method, the intensity information in a patch is converted into a LBT which is then combined with the labels of corresponding patches from the atlas to form an atom of a dictionary. The initial labels of target images are estimated through a rough segmentation. The voxel in a patch to be labelled is also constructed as a vector similar to the atom. The voxel vector is then modelled as a sparse linear combination of the atoms in the dictionary. Experimental results on two MR brain data sets demonstrated that the proposed method is efficient in the segmentation which can achieve competitive performance compared with the state-of-the-art methods.

Automatic colorization using fully convolutional networks

We propose an approach for automatically colorizing grayscale images using fully convolutional networks (FCNs). In contrast to traditional colorization methods, our approach operates only on grayscale images without any manual assistance. We first build an end-to-end deep learning network based on an FCN. Global, midlevel, and local features are extracted from the network and fused to construct each deconvolutional layer. To ensure color consistency, a low-frequency regularization term is presented to maintain the coherence between neighboring pixels. We then present an improved level-set method, which we apply to the output of the FCN to repair color bleeding caused by the rough segmentation performed by the FCN. To evaluate our approach, we compare the objective image quality resulting from our method with the results of other methods by assessing the peak signal-to-noise ratio, the mean squared error, the structural similarity index (SSIM), and the multiscale SSIM (MS-SSIM). In addition, we design a Turing test to evaluate the subjective image quality. The results show that our colorized images more closely resemble the ground-truth images and are more robust than those produced via other methods.

A level set image segmentation method based on a cloud model as the priori contour

A novel image segmentation method combining a cloud model and a level set (CM-LS) is proposed in this article. At present, the cloud model can only obtain the rough segmentation result of an image, but the level set method is sensitive to the initial contour. The core idea of this method is to use the rough segmentation result of cloud model as the initial contour of the level set and then obtain the final result by the contour evolution. In this method, the cloud model is used to decompose the boundary of the image, which reduces the occurrence probability and occurrence degree of the instability problem caused by artificial intervention; at the same time, the convergence of the level set function is accelerated, and the initializing operation of the level set function that uses the cloud model algorithm can also effectively reduce the noise sensitivity of the function itself. Compared with the conventional level set method, the proposed method is general and accurate. The experimental data set in this article includes natural images of the Berkeley database, medical images and synthetic noise images. The experimental results show that the method is effective.

Hybrid two-stage active contour method with region and edge information for intensity inhomogeneous image segmentation.

This paper presents a novel two-stage image segmentation method using an edge scaled energy functional based on local and global information for intensity inhomogeneous image segmentation. In the first stage, we integrate global intensity term with a geodesic edge term, which produces a preliminary rough segmentation result. Thereafter, by taking final contour of the first stage as initial contour, we begin second stage segmentation process by integrating local intensity term with geodesic edge term to get final segmentation result. Due to the suitable initialization from the first stage, the second stage precisely achieves desirable segmentation result for inhomogeneous image segmentation. Two stage segmentation technique not only increases the accuracy but also eliminates the problem of initial contour existed in traditional local segmentation methods. The energy function of the proposed method uses both global and local terms incorporated with compacted geodesic edge term in an additive fashion which uses image gradient information to delineate obscured boundaries of objects inside an image. A Gaussian kernel is adapted for the regularization of the level set function and to avoid an expensive re-initialization. The experiments were carried out on synthetic and real images. Quantitative validations were performed on Multimodal Brain Tumor Image Segmentation Benchmark (BRATS) 2015 and PH2 skin lesion database. The visual and quantitative comparisons will demonstrate the efficiency of the proposed method.

Classify-normalize-classify

Changing atmospheric conditions often result in a data distribution shift in remote sensing images for different dates and locations making it difficult to discriminate between various classes of interest. To alleviate this data shift issue, we introduce a novel supervised classification framework, called Classify-Normalize-Classify (CNC). The proposed scheme uses a two classifier approach where the first classifier performs a rough segmentation of the class of interest (COI) in the input image. Then, the median signal of the estimated COI regions is subtracted from all image pixels values to normalize them. Finally, the second classifier is applied to the normalized image to produce the refined COI segmentation. The proposed methodology was tested to detect deforestation using bitemporal Landsat 8 OLI images over the Amazon rainforest. The CNC framework compared favorably to benchmark masks of the PRODES program and state-of-the-art classifiers run on surface reflectance images provided by USGS.

A learning-based visual saliency prediction model for stereoscopic 3D video (LBVS-3D)

Saliency prediction models provide a probabilistic map of relative likelihood of an image or video region to attract the attention of the human visual system. Over the past decade, many computational saliency prediction models have been proposed for 2D images and videos. Considering that the human visual system has evolved in a natural 3D environment, it is only natural to want to design visual attention models for 3D content. Existing monocular saliency models are not able to accurately predict the attentive regions when applied to 3D image/video content, as they do not incorporate depth information. This paper explores stereoscopic video saliency prediction by exploiting both low-level attributes such as brightness, color, texture, orientation, motion, and depth, as well as high-level cues such as face, person, vehicle, animal, text, and horizon. Our model starts with a rough segmentation and quantifies several intuitive observations such as the effects of visual discomfort level, depth abruptness, motion acceleration, elements of surprise, size and compactness of the salient regions, and emphasizing only a few salient objects in a scene. A new fovea-based model of spatial distance between the image regions is adopted for considering local and global feature calculations. To efficiently fuse the conspicuity maps generated by our method to one single saliency map that is highly correlated with the eye-fixation data, a random forest based algorithm is utilized. The performance of the proposed saliency model is evaluated against the results of an eye-tracking experiment, which involved 24 subjects and an in-house database of 61 captured stereoscopic videos. Our stereo video database as well as the eye-tracking data are publicly available along with this paper. Experiment results show that the proposed saliency prediction method achieves competitive performance compared to the state-of-the-art approaches.

Soft fuzzy rough set-based MR brain image segmentation

Fuzzy sets, rough sets are efficient tools to handle uncertainty and vagueness in the medical images and are widely used for medical image segmentation. Soft sets are a new mathematical approach to uncertainty and vagueness. In this paper, a hybrid segmentation algorithm based on soft sets namely soft fuzzy rough c-means is proposed to extract the white matter, gray matter and the cerebro spinal fluid from MR brain image with bias field correction. In this algorithm, soft fuzzy rough approximations are applied to obtain the rough regions of image. These approximations are free from defining thresholds, weight parameters and are less complex compared to the existing rough set based algorithms. Soft sets use similarity coefficients to find the similarity of the clusters formed in present and previous step. The proposed algorithm does not involve any negative region, hence all the pixels participate in clustering avoiding clustering mistakes. Also, the histogram based centroids choose the centroids close to the ground truth that in turn effect the definition of approximations, standardizing the clusters. The proposed algorithm evaluated through simulation, compared it with existing k-means, rough k-means, fuzzy c-means and other hybrid algorithms. The soft fuzzy rough c-means algorithm outperforms the considered algorithms in all analyzed scenarios even in extracting the tumor from the brain tissue.

Template based rodent brain extraction and atlas mapping.

Accurate rodent brain extraction is the basic step for many translational studies using MR imaging. This paper presents a template based approach with multi-expert refinement to automatic rodent brain extraction. We first build the brain appearance model based on the learning exemplars. Together with the template matching, we encode the rodent brain position into the search space to reliably locate the rodent brain and estimate the rough segmentation. With the initial mask, a level-set segmentation and a mask-based template learning are implemented further to the brain region. The multi-expert fusion is used to generate a new mask. We finally combine the region growing based on the histogram distribution learning to delineate the final brain mask. A high-resolution rodent atlas is used to illustrate that the segmented low resolution anatomic image can be well mapped to the atlas. Tested on a public data set, all brains are located reliably and we achieve the mean Jaccard similarity score at 94.99% for brain segmentation, which is a statistically significant improvement compared to two other rodent brain extraction methods.

A static video summarization method based on the sparse coding of features and representativeness of frames

This paper presents a video summarization method that is specifically for the static summary of consumer videos. Considering that the consumer videos usually have unclear shot boundaries and many low-quality or meaningless frames, we propose a two-step approach where the first step skims a video and the second step performs content-aware clustering with keyframe selection. Specifically, the first step removes most of redundant frames that contain only little new information by employing the spectral clustering method with color histogram features. As a result, we obtain a condensed video that is shorter and has clearer temporal boundaries than the original. In the second step, we perform rough temporal segmentation and then apply refined clustering for each of the temporal segments, where each frame is represented by the sparse coding of SIFT features. The keyframe selection from each cluster is based on the measure of representativeness and visual quality of frames, where the representativeness is defined from the sparse coding and the visual quality is the combination of contrast, blur, and image skew measures. The problem of keyframe selection is to find the frames that have both representativeness and high quality, which is formulated as an optimization problem. Experiments on videos with various lengths show that the resulting summaries closely follow the important contents of videos.

Learning to segment with image-level annotations

Recently, deep convolutional neural networks (DCNNs) have significantly promoted the development of semantic image segmentation. However, previous works on learning the segmentation network often rely on a large number of ground-truths with pixel-level annotations, which usually require considerable human effort. In this paper, we explore a more challenging problem by learning to segment under image-level annotations. Specifically, our framework consists of two components. First, reliable hypotheses based localization maps are generated by incorporating the hypotheses-aware classification and cross-image contextual refinement. Second, the segmentation network can be trained in a supervised manner by these generated localization maps. We explore two network training strategies for achieving good segmentation performance. For the first strategy, a novel multi-label cross-entropy loss is proposed to train the network by directly using multiple localization maps for all classes, where each pixel contributes to each class with different weights. For the second strategy, the rough segmentation mask can be inferred from the localization maps, and then the network is optimized based on the single-label cross-entropy loss with the produced masks. We evaluate our methods on the PASCAL VOC 2012 segmentation benchmark. Extensive experimental results demonstrate the effectiveness of the proposed methods compared with the state-of-the-arts.

Automatic Approach for Lung Segmentation with Juxta-Pleural Nodules from Thoracic CT Based on Contour Tracing and Correction.

This paper presents a fully automatic framework for lung segmentation, in which juxta-pleural nodule problem is brought into strong focus. The proposed scheme consists of three phases: skin boundary detection, rough segmentation of lung contour, and pulmonary parenchyma refinement. Firstly, chest skin boundary is extracted through image aligning, morphology operation, and connective region analysis. Secondly, diagonal-based border tracing is implemented for lung contour segmentation, with maximum cost path algorithm used for separating the left and right lungs. Finally, by arc-based border smoothing and concave-based border correction, the refined pulmonary parenchyma is obtained. The proposed scheme is evaluated on 45 volumes of chest scans, with volume difference (VD) 11.15 ± 69.63 cm3, volume overlap error (VOE) 3.5057 ± 1.3719%, average surface distance (ASD) 0.7917 ± 0.2741 mm, root mean square distance (RMSD) 1.6957 ± 0.6568 mm, maximum symmetric absolute surface distance (MSD) 21.3430 ± 8.1743 mm, and average time-cost 2 seconds per image. The preliminary results on accuracy and complexity prove that our scheme is a promising tool for lung segmentation with juxta-pleural nodules.

Shape-intensity prior level set combining probabilistic atlas and probability map constrains for automatic liver segmentation from abdominal CT images.

Propose a fully automatic 3D segmentation framework to segment liver on challenging cases that contain the low contrast of adjacent organs and the presence of pathologies from abdominal CT images. First, all of the atlases are weighted in the selected training datasets by calculating the similarities between the atlases and the test image to dynamically generate a subject-specific probabilistic atlas for the test image. The most likely liver region of the test image is further determined based on the generated atlas. A rough segmentation is obtained by a maximum a posteriori classification of probability map, and the final liver segmentation is produced by a shape-intensity prior level set in the most likely liver region. Our method is evaluated and demonstrated on 25 test CT datasets from our partner site, and its results are compared with two state-of-the-art liver segmentation methods. Moreover, our performance results on 10 MICCAI test datasets are submitted to the organizers for comparison with the other automatic algorithms. Using the 25 test CT datasets, average symmetric surface distance is [Formula: see text]mm (range 0.62-2.12mm), root mean square symmetric surface distance error is [Formula: see text]mm (range 0.97-3.01mm), and maximum symmetric surface distance error is [Formula: see text]mm (range 12.73-26.67mm) by our method. Our method on 10 MICCAI test data sets ranks 10th in all the 47 automatic algorithms on the site as of July 2015. Quantitative results, as well as qualitative comparisons of segmentations, indicate that our method is a promising tool to improve the efficiency of both techniques. The applicability of the proposed method to some challenging clinical problems and the segmentation of the liver are demonstrated with good results on both quantitative and qualitative experimentations. This study suggests that the proposed framework can be good enough to replace the time-consuming and tedious slice-by-slice manual segmentation approach.

DCT와 Guided 필터를 이용한 실시간 영상 분류

In this paper, we present a novel segmentation method that can extract new foreground objects from a current frame in real-time. It is performed by detecting differences between the current frame and reference frame taken from a fixed camera. We minimize computing complexity for real-time video processing. First DCT (Discrete Cosine Transform) is utilized to generate rough binary segmentation maps where foreground and background regions are separated. DCT shows better result of texture analysis than previous methods where texture analysis is performed in spatial domain. It is because texture analysis in frequency domain is easier than that in special domain and intensity and texture in DCT are taken into account at the same time. We maximize run-time efficiency of DCT by considering color information to analyze object region prior to DCT process. Last we use Guided filter for natural matting of the generated binary segmentation map. In general, Guided filter can enhance quality of intermediate result by incorporating guidance information. However, it shows some limitations in homogeneous area. Therefore, we present an additional method which can overcome them. Keyword : DCT, matting, segmentation, Guided Filter a) 경희대학교 전자정보대학 (Kyung Hee University) b) kt 융합기술원 (kt fusion technology institute) ‡Corresponding Author : 김원하(Wonha Kim) E-mail: hjk1017@ssu.ac.kr Tel: +82-2-826-9063 ※ 본 연구는 산업통상지원부의 재원으로 산업핵심기술개발사업의 지원을 받아 수행 되었습니다.(10052967, 재난재해 대응용 특수목적기계 통합제어시스 템 개발) ・Manuscript received August 17, 2015; revised September 22, 2015; accepted September 22, 2015. 특집논문 (Special Paper) 방송공학회논문지 제20권 제5호, 2015년 9월 (JBE Vol. 20, No. 5, September 2015) http://dx.doi.org/10.5909/JBE.2015.20.5.718 ISSN 2287-9137 (Online) ISSN 1226-7953 (Print) 신현학 외 2인 : DCT와 Guided 필터를 이용한 실시간 영상 분류 719 (Hyunhak Shin et al.: Video Segmentation Using DCT and Guided Filter in real time)

Kidney segmentation in CT sequences using SKFCM and improved GrowCut algorithm.

BackgroundOrgan segmentation is an important step in computer-aided diagnosis and pathology detection. Accurate kidney segmentation in abdominal computed tomography (CT) sequences is an essential and crucial task for surgical planning and navigation in kidney tumor ablation. However, kidney segmentation in CT is a substantially challenging work because the intensity values of kidney parenchyma are similar to those of adjacent structures.ResultsIn this paper, a coarse-to-fine method was applied to segment kidney from CT images, which consists two stages including rough segmentation and refined segmentation. The rough segmentation is based on a kernel fuzzy C-means algorithm with spatial information (SKFCM) algorithm and the refined segmentation is implemented with improved GrowCut (IGC) algorithm. The SKFCM algorithm introduces a kernel function and spatial constraint into fuzzy c-means clustering (FCM) algorithm. The IGC algorithm makes good use of the continuity of CT sequences in space which can automatically generate the seed labels and improve the efficiency of segmentation. The experimental results performed on the whole dataset of abdominal CT images have shown that the proposed method is accurate and efficient. The method provides a sensitivity of 95.46% with specificity of 99.82% and performs better than other related methods.ConclusionsOur method achieves high accuracy in kidney segmentation and considerably reduces the time and labor required for contour delineation. In addition, the method can be expanded to 3D segmentation directly without modification.

Enhanced Crack Segmentation Algorithm Using 3D Pavement Data

AbstractAutomatic pavement crack segmentation has gained attention among researchers and transportation agencies over the past two decades. However, most existing algorithms using two-dimensional (2D) pavement intensity images cannot provide a satisfactory performance. With the advent of sensing technology, three-dimensional (3D) line laser pavement imaging systems have become available. The objective of this paper is to propose an enhanced dynamic optimization algorithm employing the advantages of 3D pavement data to improve crack segmentation. The proposed algorithm consists of three major stages. First, a two-step Gaussian filter is applied to remove outliers from the collected laser data and rectify the profile in order to reduce the influence of cross-slope and ruts on crack segmentation. Then, a rough crack segmentation stage is conducted to adaptively identify the crack regions of interest. Finally, a bounding box and major orientation for each valid crack region of interest will provide searching ...

Semiautomatic detection of villi in confocal endoscopy for the evaluation of celiac disease.

Celiac Disease (CD) is an immune-mediated enteropathy, diagnosed in the clinical practice by intestinal biopsy and the concomitant presence of a positive celiac serology. Confocal Laser Endomicroscopy (CLE) allows skilled and trained experts to potentially perform in vivo virtual histology of small-bowel mucosa. In particular, it allows the qualitative evaluation of mucosa alteration such as a decrease in goblet cells density, presence of villous atrophy or crypt hypertrophy. We present a semi-automatic method for villi detection from confocal endoscopy images, whose appearance change in case of villous atrophy. Starting from a set of manual seeds, a first rough segmentation of the villi is obtained by means of mathematical morphology operations. A merge and split procedure is then performed, to ensure that each seed originates a different region in the final segmentation. A border refinement process is finally performed, evolving the shape of each region according to local gradient intensities. Mean and median Dice coefficients for 290 villi originating from 66 images when compared to manually obtained ground truth are 80.71% and 87.96% respectively.

An automated pulmonary parenchyma segmentation method based on an improved region growing algorithmin PET-CT imaging

To address the incomplete problem in pulmonary parenchyma segmentation based on the traditional methods, a novel automated segmentation method based on an eight-neighbor region growing algorithm with left-right scanning and four-corner rotating and scanning is proposed in this paper. The proposed method consists of four main stages: image binarization, rough segmentation of lung, image denoising and lung contour refining. First, the binarization of images is done and the regions of interest are extracted. After that, the rough segmentation of lung is performed through a general region growing method. Then the improved eight-neighbor region growing is used to remove noise for the upper, middle, and bottom region of lung. Finally, corrosion and expansion operations are utilized to smooth the lung boundary. The proposed method was validated on chest positron emission tomography-computed tomography (PET-CT) data of 30 cases from a hospital in Shanxi, China. Experimental results show that our method can achieve an average volume overlap ratio of 96.21 ± 0.39% with the manual segmentation results. Compared with the existing methods, the proposed algorithm segments the lung in PET-CT images more efficiently and accurately.

Variants of seeded region growing

Seeded region growing (SRG) is a fast, effective and robust method for image segmentation. It begins with placing a set of seeds in the image to be segmented, where each seed could be a single pixel or a set of connected pixels. Then SRG grows these seeds into regions by successively adding neighbouring pixels to them. It finishes when all pixels in the image are assigned to one (and only one) region. The growing strategy of SRG implicitly assumes that pixels from the same region share the same greyvalue. As a first contribution, this study develops new modifications to SRG so that this constant greyvalue assumption is relaxed. Since the growing strategy of SRG does not impose any constraints or restrictions on the shapes of the growing regions, quite often SRG would produce very rough segmentation boundaries even the true boundaries were smooth. The second contribution of this study is the proposal of a stabilised SRG that encourages smoother boundaries and prevents the so-called leakage problem. All these new variants are conceptually simple and easy to implement. They are tested with simulated and real images, and are shown to have better performances than the original SRG.

Robust surface segmentation and edge feature lines extraction from fractured fragments of relics

Abstract Surface segmentation and edge feature lines extraction from fractured fragments of relics are essential steps for computer assisted restoration of fragmented relics. As these fragments were heavily eroded, it is a challenging work to segment surface and extract edge feature lines. This paper presents a novel method to segment surface and extract edge feature lines from triangular meshes of irregular fractured fragments. Firstly, a rough surface segmentation is accomplished by using a clustering algorithm based on the vertex normal vector. Secondly, in order to differentiate between original and fracture faces, a novel integral invariant is introduced to compute the surface roughness. Thirdly, an accurate surface segmentation is implemented by merging faces based on face normal vector and roughness. Finally, edge feature lines are extracted based on the surface segmentation. Some experiments are made and analyzed, and the results show that our method can achieve surface segmentation and edge extraction effectively.