Performance Analysis of Otsu-Based Thresholding Algorithms: A Comparative Study

Image thresholding is a widely used technology for a lot of computer vision applications, and among various global thresholding algorithms, Otsu-based approaches are very popular due to their simplicity and effectiveness. While the usage of Otsu-based thresholding methods is well discussed, the performance analyses of these methods are rather limited. In this paper, we first review nine Otsu-based approaches and categorize them based on their objective functions, preprocessing, and postprocessing strategies. Second, we conduct several experiments to analyze the model characteristics using different scene parameters both on synthetic images and real-world cell images. We put more attention to examine the variance of foreground object and the effect of the distance between mean values of foreground and background. Third, we explore the robustness of algorithms by introducing two typical kinds of noises under different intensities and compare the running time of each method. Experimental results show that NVE, WOV, and Xing’s methods are more robust to the distance of mean values of foreground and background. The large foreground variance will cause a larger threshold value. Experiments on cell images show that foreground miss detection becomes serious when the intensities of foreground pixels change drastically. We conclude that almost all algorithms are significantly affected by Salt&Pepper and Gaussian noises. Interestingly, we find that ME increases almost linearly with the intensity of Salt&Pepper noise. In terms of algorithms’ time cost, methods with no preprocessing and postprocessing steps have more advantages. All these findings can serve as a guideline for image thresholding when using Otsu-based thresholding approaches.

Image segmentation and analysis for densification mapping of nanoporous gold after nanoindentation

Segmentation of scanning electron microscopy (SEM) images of focused ion beam (FIB) cross-sections through indented regions in nanoporous gold (np-Au) is carried out. A key challenge for image analysis of open porous materials is the appropriate binarization of the pore and gold ligament regions while excluding material lying below the cross-sectional plane. Here, a manual approach to thresholding is compared to global and local approaches. The global thresholding resulted in excessive deviations from the nominal solid fraction, due to a strong gray-scale gradient caused by the tilt angle during imaging and material shadowing. In contrast, the local thresholding approach delivered local solid fractions that were free of global gradients, and delivered a quality comparable to the manual segmentation. The extracted densification profiles vertically below the indenter as well as parallel to the surface showed an exponential-type decay from the indenter tip towards the nominal value of 1 far from the indenter. Graphic abstract

Proposta de Implementação em Hardware do Algoritmo de Otsu Aplicado ao Rastreamento em Tempo Real de Vermes

This work proposes an implementation in Field Programmable GateArray (FPGA) of the Otsu’s method applied to real-time trackingof worms called Caenorhabditis elegans. Real-time tracking is necessaryto measure changes in the worm’s behavior in response totreatment with Ribonucleic Acid (RNA) interference. Otsu’s methodis a global thresholding algorithm used to define an optimal thresholdbetween two classes. However, this technique in real-time applicationsassociated with the processing of high-resolution videoshas a high computational cost because of the massive amount ofdata generated. Otsu’s algorithm needs to identify the worms ineach frame captured by a high-resolution camera in a real-timeanalysis of the worm’s behavior. Thus, this work proposes a highperformanceimplementation of Otsu’s algorithm in FPGA. Theresults show it was possible to achieve a speedup up to 5 timeshigher than similar works in the literature.

CLOUD IDENTIFICATION FROM MULTITEMPORAL LANDSAT-8 USING K-MEANS CLUSTERING

In the processing and analysis of remote-sensing data, cloud that interferes with earth-surface data is still a challenge. Many methods have already been developed to identify cloud, and these can be classified into two categories: single-date and multi-date identification. Most of these methods also utilize the thresholding method which itself can be divided into two categories: local thresholding and global thresholding. Local thresholding works locally and is different for each pixel, while global thresholding works similarly for every pixel. To determine the global threshold, two approaches are commonly used: fixed value as threshold and adapted threshold. In this paper, we propose a cloud-identification method with an adapted threshold using K-means clustering. Each related multitemporal pixel is processed using K-means clustering to find the threshold. The threshold is then used to distinguish clouds from non-clouds. By using the L8 Biome cloud-cover assessment as a reference, the proposed method results in Kappa coefficient of above 0.9. Furthermore, the proposed method has lower levels of false negatives and omission errors than the FMask method.