China
Africa
Explore chapters and articles related to this topic
Image Segmentation
Published in Vipin Tyagi, Understanding Digital Image Processing, 2018
Thresholding is a commonly used method for image segmentation because of its simple implementation, low computation cost and intuitive properties. The thresholding-based image segmentation method divides pixels on the basis of intensity level or gray level of an image. Hence, this method divides the whole image into background and foreground regions where it is assumed that objects (foreground regions) have pixels of greater intensity levels as compared to the pixels in the background. Due to the aforementioned reason, thresholding-based methods are applicable where the objects in the image differ in their gray level distribution. In thresholding-based methods, all the pixels belonging to an object are given a value of “1” while background pixels are given value of “0”. Therefore, a binary image (image with pixel values 0 or 1) is generated.
Basics of Image Processing
Published in Maheshkumar H. Kolekar, Intelligent Video Surveillance Systems, 2018
Algorithm 4 is used to obtain threshold T automatically. In general, if T0 $ T_{0} $ is larger, the algorithm will perform fewer iterations. The initial threshold T can be selected as average intensity of the image. Global thresholding uses a fixed threshold for all pixels in the image. Hence, it works only if the intensity histogram of the given image contains two separate peaks corresponding to the objects and background. As shown in Figure 1.21(b), global thresholding failed to segment the river and mountain on the backgorund of the image. Hence, to achieve better segmentaion performance, a local thresholding technique is used.
Feature Extraction with Statistics and Decision Science Algorithms
Published in Ni-Bin Chang, Kaixu Bai, Multisensor Data Fusion and Machine Learning for Environmental Remote Sensing, 2018
Thresholding techniques work mainly by relying on spectral differences between various targets to extract the desired features; hence the threshold values for the same target could even vary between images due to radiometric distortions caused by various factors such as illumination conditions. To account for such drawbacks, more flexible approaches should be applied. In image interpretation, the shapes of targets are often considered good features for further pattern recognition, since the perimeter of an object can be easily perceived by human vision. Hence, detecting the shape features from a given imagery is critical to the subsequent feature extraction. Essentially, the shape of an object is commonly treated as a step change in the intensity levels (Nixon and Aguado, 2012).
Enhanced fuzzy Gaussian networks for sputum image based Mycobacterium detection
Published in The Imaging Science Journal, 2023
Let be the database and the number of sputum smear microscopic images. The sputum smear image database is given in Equation (1) as where represent the sputum smear microscopic image. Each image in the database is considered to be of size . The segmentation process is performed by the colour space (CS) transformation model and Otsu thresholding. In Otsu thresholding, the threshold value is set based on the variance computed at each intensity level of the image. The maximum variance is set as a threshold using weight and mean values of background and foreground pixel values [13]. Pseudocode 1 represents the steps carried out for the segmentation process of images in the CS model.
TaylorCSROA-based Deep Residual Network: An Optimization driven Deep Network for the multilevel spinal cord disease classification
Published in Computer Methods in Biomechanics and Biomedical Engineering: Imaging & Visualization, 2023
The initial process in the multilevel spinal cord disease classification process is adaptive thresholding. For adaptive thresholding, the input MR image from the data set is considered. The adaptive thresholding method is developed by modifying Wellner’s method (Wellner 1993). In the adaptive thresholding, the individual pixel of the image is compared with the surrounding pixel’s mean. Particularly, while crossing the image, the moving average of certain pixels is obtained. The variation in the soft gradient is neglected, and hard contrast lines are protected by comparing the pixel and neighbouring pixel. The advantage of the adaptive thresholding method is that it only requires a single pass throughout the image. The adaptive thresholding method produced an independent output and does not require the type of processing technique implemented for image processing. The spinal cord segments obtained from the adaptive thresholding are provided to the disc localisation module for the disc identification.
Failure patterns of solder joints identified through lifetime vibration tests
Published in Nondestructive Testing and Evaluation, 2023
Kangkana Baishya, David M. Harvey, Teresa Partida Manzanera, Guangming Zhang, Derek R. Braden
Once the intensity/gain mismatches were corrected as illustrated in section 5.1, MATLAB coding was used to select individual solder joints for degradation analysis. Then, the first step to carry out the analysis was thresholding. Thresholding is a simple but effective technique of image segmentation. For most images, in general, the grey levels of pixels belonging to an object are substantially different from the grey level of the background. An appropriate threshold value can be set to differentiate the object and the background. For example, any pixels with values either greater than or less than a threshold value are treated as the main object and the rest are considered as background. The most important parameter in a thresholding technique is the decision of the threshold value. Many features in the image can be used to set the threshold parameter, for example, histogram [22,23], and gradient information [24].