Otsu's Global Thresholding Research Articles

Estimating ground surface visibility on thermal images from drone wildlife surveys in forests

A thermal drone – an unmanned aerial vehicle equipped with a thermal camera – is a new tool in wildlife research. It enables effective surveys of animal species difficult to monitor with conventional methods. Over open terrain the detection of warm-blooded large animals with a thermal drone is nearly certain. However, in forest surveys, individuals hidden under the tree canopy may remain undetected which can lead to biased estimates of population size. Thus, it is important to estimate the percentage of surveyed area not obscured by tree canopy where animals can be easily detected. This figure may be used as a simple correction factor for animal counts from thermal images (thermograms) or as one of the covariates in more advanced statistical models of population density. We tested four methods of automatic image segmentation: 1) Otsu global thresholding, 2) local thresholding with Gaussian kernel, 3) Chan-Vese active contour model without edges, and 4) logistic regression and assessed each of their performances in differentiating between canopy and forest floor on a sample of 100 thermograms acquired with a drone during wildlife surveys in a range of forest habitats. We used segmentation results to calculate the estimated percentage of visible ground surface on images and compared it to the percentages assessed by manual classification. The Otsu global thresholding gave results that were in best agreement with the manual classification, had the fastest computing time and was superior to the other three segmentation methods. Visual inspection of the Otsu segmentation results showed that this method accurately separates tree canopy and ground surface on thermograms obtained across all forest types as well as in partially open areas. However, landscape features such as roads or water bodies that have a temperature similar to that of trees may cause segmentation errors and lead to underestimation of the ground visibility. While the use of Otsu segmentation to automatically assess percentage of visible ground surface on thermal images from drone surveys has certain limitations, the method is fast and can be readily applied by wildlife researchers. This approach can aid in correcting bias in estimates of animal population density and contribute to a better understanding of the use of thermal drones for wildlife surveys in forests.

Virtual characterisation of porcupine quills using X-ray micro-CT

ABSTRACT Morphological properties and structures of the porcupine quills of different sizes can be explored in detail using micro-CT scans. Information including the porosity, volume thickness, number and lengths of stiffeners within quills are extracted via two modalities, using commercially available software and custom-developed scripts. Three segmentation methods, including the Otsu global thresholding, histographic segmentation and deep learning segmentation, are applied to identify the porosity variation based on different segmentation methods. Over segmentation is found when applying Otsu global thresholding, yielded the highest porosity. Histographic segmentation performed better than Otsu thresholding in segmenting CT slices, however speckles are observed in foam areas with lower intensity. Among all methods, deep learning segmentation resulted in the most reliable segmentation, and showed a consistent porosity across different quill sizes. Obtained results provide essential information for biomimicry research with an aim towards designing stronger and lighter structures for various engineering applications.

FWENet: a deep convolutional neural network for flood water body extraction based on SAR images

ABSTRACT As one of the most severe natural disasters in the world, floods caused substantial economic losses and casualties every year. Timely and accurate acquisition of flood inundation extent could provide technical support for relevant departments in the field of flood emergency response and disaster relief. Given the accuracy of existing research works extracting flood inundation extent based on Synthetic Aperture Radar (SAR) images and deep learning methods is relatively low, this study utilized Sentinel-1 SAR images as the data source and proposed a novel model named flood water body extraction convolutional neural network (FWENet) for flood information extraction. Then three classical semantic segmentation models (UNet, Deeplab v3 and UNet++) and two traditional water body extraction methods (Otsu global thresholding method and Object-Oriented method) were compared with the FWENet model. Furthermore, this paper analyzed the water body area change situations of Poyang Lake. The main results of this paper were as follows: Compared with other five water body extraction methods, the FWENet model achieved the highest water body extraction accuracy, its F1 score and mean intersection over union (mIoU) were 0.9871 and 0.9808, respectively. This study could guarantee the subsequent research on flood extraction based on SAR images.

Frequency based edge-texture feature using Otsu’s based enhanced local ternary pattern technique for digital image splicing detection

Sharing information through images is a trend nowadays. Advancements in the technology and user-friendly image editing tool make easy to edit the image and spread fake news through different social networking platforms. Forged image has been generated through an advanced image editing tool, so it is very challenging for image forensics to detect the micro discrepancy which distorted the micro pattern. This paper proposes an image forensic detection technique, which implies multi-level discrete wavelet transform to implement digital image filtering. Canny edge detection technique is implemented to detect the edge of the image to implement Otsu’s based enhanced local ternary pattern (OELTP), which can detect forgery-related artifact. DWT is implemented over Cb and Cr components of the image and using edge texture to improve the Otsu global threshold, which is used to extract features using ELTP technique. Support vector machine (SVM) is used for classification to find the image is forged or not. The performance of the work evaluated on three different open available data sets CASIA v1, CASIA v2, and Columbia. Our proposed work gives better results with some of the previous states of the work in terms of detection accuracy.

Exploration and evaluation of efficient pre-processing and segmentation technique for breast cancer diagnosis based on mammograms

Breast cancer is the second leading cause of death for women everywhere in the world. Since the reason behind the disease remains unknown, early detection and diagnosis is the key challenge for breast cancer control. In this work, mammogram images are initially subject to pre-processing using Laplacian filter for enhancement of tumour regions, Gaussian mixture model, Gaussian kernel FCM, Otsu global thresholding and FCM technique are employed for segmentation. Further, the efficiency of segmentation techniques is analyzed by classifying the samples into benign, malignant and healthy using Gray Level Co-occurrence Matrix (GLCM) features. Linear discriminant analysis classifier is used a combination based on which efficiency used for classification of mammograms. Ensemble methods are evaluated. The efficiency has resulted in better accuracy with the ensemble-based method. The experimentation is conducted in the mini MIAS database of mammograms, and the efficiency of the linear discriminant analyzer is found to be 89.19% for GKFCM, 83.78% with Otsu and 78.38% with FCM method with GLCM features.

Morphological image processing for multiscale analysis of super-resolution ultrasound images of tissue microvascular networks.

Diabetes is a major disease and known to impair microvascular recruitment due to insulin resistance. Previous quantifications of the changes in microvascular networks at the capillary level were being performed with either full or manually selected region-of-interests (ROIs) from super-resolution ultrasound (SR-US) images. However, these approaches were imprecise, time-consuming, and unsuitable for automated processes. Here we provided a custom software solution for automated multiscale analysis of SR-US images of tissue microvascularity patterns. An Acuson Sequoia 512 ultrasound (US) scanner equipped with a 15L8-S linear array transducer was used in a nonlinear imaging mode to collect all data. C57BL/6J male mice fed standard chow and studied at age 13-16 wk comprised the lean group (N = 14), and 24-31 wk-old mice who received a high-fat diet provided the obese group (N = 8). After administration of a microbubble (MB) contrast agent, the proximal hindlimb adductor muscle of each animal was imaged (dynamic contrast-enhanced US, DCE-US) for 10 min at baseline and again at 1 h and towards the end of a 2 h hyperinsulinemic-euglycemic clamp. Vascular structures were enhanced with a multiscale vessel enhancement filter and binary vessel segments were delineated using Otsu's global threshold method. We then computed vessel diameters by employing morphological image processing methods for quantitative analysis. Our custom software enabled automated multiscale image examination by defining a diameter threshold to limit the analysis at the capillary level. Longitudinal changes in AUC, IPK, and MVD were significant for lean group (p < 0.02 using Full-ROI and p < 0.01 using 150 μm-ROI) and for obese group (p < 0.02 using Full-ROI, p < 0.03 using 150 μm-ROI). By eliminating large vessels from the ROI (above 150 μm in diameter), perfusion parameters were more sensitive to changes exhibited by the smaller vessels, that are known to be more impacted by disease and treatment.

A Simple Approach to Automated Brain Tumor Segmentation and Classification

Brain tumor is the abnormal growth of superfluous cells in central nervous system or brain. It is fact that brain tumor is second most common of cancer death among young people. There are two key categories of brain tumor as cancerous and non cancerous. The cancerous brain tumor is called as Malignant. It spreads very quickly and difficult to remove. The non-cancerous tumor, called Benign, growth rate is very slow as compared to malignant one and easy to remove. The work proposes a simple but more efficient method to detect and segment the brain tumor from the MRI image. The proposed work based on the threshold segmentation for the segmentation of the brain tumor. The MRI image of the brain is taken and processed in such a way so that the tumor is extracted from the given MRI image and displays the segmented part of the image which contains the tumor. The otsu global threshold performs tumor segmentation and image area opening applies to remove the small components form the tumor portion. The gray level co-occurrence matrix and other image quality measures computes (extracts) the features from the segmented image. Support vector machine classifier is finally classifies the tumor, either benign or malignant, based on the extracted features.

ECTG: an automatic procedure to extract digital cardiotocographic signals from digital images

Background and objectiveCardiotocography (CTG), consisting in the simultaneous recording of fetal heart rate (FHR) and maternal uterine contractions (UC), is a popular clinical test to assess fetal health status. Typically, CTG machines provide paper reports that are visually interpreted by clinicians. Consequently, visual CTG interpretation depends on clinician's experience and has a poor reproducibility. The lack of databases containing digital CTG signals has limited number and importance of retrospective studies finalized to set up procedures for automatic CTG analysis that could contrast visual CTG interpretation subjectivity. In order to help overcoming this problem, this study proposes an electronic procedure, termed eCTG, to extract digital CTG signals from digital CTG images, possibly obtainable by scanning paper CTG reports. MethodseCTG was specifically designed to extract digital CTG signals from digital CTG images. It includes four main steps: pre-processing, Otsu's global thresholding, signal extraction and signal calibration. Its validation was performed by means of the “CTU-UHB Intrapartum Cardiotocography Database” by Physionet, that contains digital signals of 552 CTG recordings. Using MATLAB, each signal was plotted and saved as a digital image that was then submitted to eCTG. Digital CTG signals extracted by eCTG were eventually compared to corresponding signals directly available in the database. Comparison occurred in terms of signal similarity (evaluated by the correlation coefficient ρ, and the mean signal error MSE) and clinical features (including FHR baseline and variability; number, amplitude and duration of tachycardia, bradycardia, acceleration and deceleration episodes; number of early, variable, late and prolonged decelerations; and UC number, amplitude, duration and period). ResultsThe value of ρ between eCTG and reference signals was 0.85 (P < 10−560) for FHR and 0.97 (P < 10−560) for UC. On average, MSE value was 0.00 for both FHR and UC. No CTG feature was found significantly different when measured in eCTG vs. reference signals. ConclusionseCTG procedure is a promising useful tool to accurately extract digital FHR and UC signals from digital CTG images.

Computerized detection of breast cancer in digital mammograms

In this paper, authors propose an efficient system for the segmentation of lesions in digital mammograms. The proposed detection system comprises three steps. In the first step, efficient pre-processing technique is developed using top-hat morphological transform, wavelet transform, and morphological opening–closing reconstruction filter followed by median filter. In the second step, threshold selection procedure is developed using a combination of Fuzzy C-means (FCM) clustering and Otsu global thresholding method. Finally, computed threshold is used to binarize the pre-processed image thereby yielding the suspicious mass regions from it. The free response operating characteristics (FROC) curve is used to evaluate the performance of the proposed method. Proposed system achieved the sensitivity of 94% at the rate of 1.30 false positives per image.

Error analysis in determining the centroids of circular objects in images

Detecting the size and/or location of circular object(s) in an image(s) has application in many areas, like, flow diagnostics, biomedical engineering, computer vision, etc. The detection accuracy of circular object(s) largely depends on the accuracy of centroiding algorithm and image preprocessing technique. In the present work, an error analysis is performed in determining the centroids of circular objects using synthetic images with eight different signal-to-noise ratios ranging from 2.7 to 17.8. In the first stage, four different centroiding algorithms, namely, Center of Mass, Weighted Center of Mass, Spath algorithm, and Hough transform, are studied and compared. The error analysis shows that Spath algorithm performs better than other algorithms. In the second stage, various image preprocessing techniques, consisting of two filters, namely, Median and Wiener, and five image segmentation methods, namely, Sobel, Prewitt, Laplacian of Gaussian (LoG) edge detector, basic global thresholding, and Otsu's global thresholding, are compared to determine the centroids of circular objects using Spath algorithm. It is found that Wiener filter plus LoG edge detector performs better than other preprocessing techniques. Real images of a calibration target (typical in flow diagnostics) and the secondary atomization of water droplets are then considered for centroids detection. These two images are preprocessed using Wiener filter plus LoG edge detector and then processed using Spath algorithm to detect the centroids of circular objects. It is observed that the results of real image of the calibration target and synthetic images are comparable. Also, based on visual inspection, the centroids detected in the real image of water droplets are found to be reasonably accurate.

Exudates and Blood Vessel Segmentation in Eye Fundus Images using the Fourier and Cosine Discrete Transforms

This paper presents a new method using discrete transforms to segment blood vessels and exudates in eye fundus color images. To obtain the desired segmentation, an illumination correction is previously done based on a homomorphic filter because of the uneven illuminance in the eye fundus image. To distinguish foreground objects from the background, we propose a super Gaussian bandpass filter in the discrete cosine transform DCT domain. These filters are applied on the green channel that contains information to segment pathologies. To segment exudates in the filtered DCT image, a gamma correction is applied to enhance foreground objects; in the resulting image, the Otsu's global threshold method is applied, after which, a masking operation over the effective area of the eye fundus image is performed to obtain the final segmentation of exudates. In the case of blood vessels, the negative of the image filtered with DCT is first calculated, then a median filter is applied to reduce noise and artifacts, followed by a gamma correction. Again, the Otsu's global threshold method for binarization, next a morphological closing operation is employed, and a masking operation gives the corresponding final segmentation. Illustrative examples taken from a free clinical database are included to demonstrate the capability of the proposed methods.

Visual Saliency Models for Text Detection in Real World.

This paper evaluates the degree of saliency of texts in natural scenes using visual saliency models. A large scale scene image database with pixel level ground truth is created for this purpose. Using this scene image database and five state-of-the-art models, visual saliency maps that represent the degree of saliency of the objects are calculated. The receiver operating characteristic curve is employed in order to evaluate the saliency of scene texts, which is calculated by visual saliency models. A visualization of the distribution of scene texts and non-texts in the space constructed by three kinds of saliency maps, which are calculated using Itti's visual saliency model with intensity, color and orientation features, is given. This visualization of distribution indicates that text characters are more salient than their non-text neighbors, and can be captured from the background. Therefore, scene texts can be extracted from the scene images. With this in mind, a new visual saliency architecture, named hierarchical visual saliency model, is proposed. Hierarchical visual saliency model is based on Itti's model and consists of two stages. In the first stage, Itti's model is used to calculate the saliency map, and Otsu's global thresholding algorithm is applied to extract the salient region that we are interested in. In the second stage, Itti's model is applied to the salient region to calculate the final saliency map. An experimental evaluation demonstrates that the proposed model outperforms Itti's model in terms of captured scene texts.