Weakly supervised semantic segmentation of leukocyte images based on class activation maps
0
Citation
12
Reference
10
Related Paper
Abstract:
Leukocytes are an essential component of the human defense system, accurate segmentation of leukocyte images is a crucial step towards automating detection. Most existing methods for leukocyte images segmentation relied on fully supervised semantic segmentation (FSSS) with extensive pixel-level annotations, which are time-consuming and labor-intensive. To address this issue, this paper proposes a weakly supervised semantic segmentation (WSSS) approach for leukocyte images utilizing improved class activation maps (CAMs). Firstly, to alleviate ambiguous boundary problem between leukocytes and background, preprocessing technique is employed to enhance the image quality. Secondly, attention mechanism is added to refine the CAMs generated by improving the matching of local and global features. Random walks, dense conditional random fields and hole filling were leveraged to obtain final pseudo-segmentation labels. Finally, a fully supervised segmentation network is trained with pseudo-segmentation labels. The method is evaluated on BCCD and TMAMD datasets. Experimental results demonstrate that by employing the pseudo segmentation annotations generated through this method can be utilized to train UNet as close as possible to FSSS. This method effectively reduces manual annotation cost while achieving WSSS of leukocyte images.Keywords:
Segmentation-based object categorization
Image segmentation is an important issue and it's also a classical problem.In this paper three kinds of image segmentation technology are intruduced.They are threshold segmentation method,Edge detection method and region segmentation method.New technology of Image segmentation is also described.Watershed algorithm based on tag is applied in this paper.From the simulation results we conclude that marker selection is the key to image segmentation algorithm.
Segmentation-based object categorization
Region growing
Range segmentation
Cite
Citations (0)
Thousands of image segmentation algorithms have been proposed so far, and new algorithms are proposed continuously. Image segmentation can be classified by the view of image segmentation application, or by the theoretical tool used in image segmentation. The main research fields including Otsu method, clustering, motion segmentation, image segmentation method based on graph theory and image segmentation method based on active contour. Natural computing algorithm has become a new hotspot of image segmentation.
Segmentation-based object categorization
Region growing
Range segmentation
Cite
Citations (0)
Region segmentation is one of the most common methods of medical image segmentation. However, it still has some disadvantages in practice: (1) Threshold choice will result in a poor performance in the image segmentation if there is little difference in the gray of an image. (2) Region-based segmentation algorithm is usually uncertain in defining the edge between the object and the background. This paper implements an improved algorithm to overcome these problems. In the algorithm, the FCM (fuzzy C-means) clustering method is used to improve the accuracy of image segmentation according to its stability. And Roberts operator is also added to compensate the deficiency in edge detection. This medical image segmentation method is a combination of region segmentation and edge segmentation, which is based on OTSU threshold segmentation, fuzzy C-means clustering and Roberts operator. Experiments show that the improved segmentation algorithm has a better performance than those traditional algorithms in the effect.
Segmentation-based object categorization
Region growing
Range segmentation
Cite
Citations (0)
Image segmentation technology is one of the important topics in the field of digital image research. However, there is no uniform standard for existing image segmentation methods, and the traditional image segmentation method is only suitable for some specific occasions. Therefore, it is very urgent to research and develop new theories and methods of image segmentation technology. Genetic algorithm is a method for calculating the optimal solution by simulating the biological evolution process in the natural selection and genetic mechanism of biological evolution. It has strong robustness, parallelism, adaptability and fast convergence. It can be applied in image segmentation technology to determine the segmentation threshold. Therefore, this paper studies the image segmentation based on genetic algorithm, and compares different image segmentation algorithms. The experimental results show that the image segmentation effect based on genetic algorithm is better than the traditional image segmentation.
Segmentation-based object categorization
Region growing
Robustness
Range segmentation
Cite
Citations (10)
A single defect image segmentation algorithm cannot obtain the desired segmentation results for all images because of the defect diversity. A parallel combination segmentation method is proposed to integrate multiple results of the different segmentation algorithms to obtain the desired defect segmentation map for high-speed aluminum surface defect detection. Two types of segmentation algorithms are designed in this combination framework, namely, the automatic threshold segmentation based on the image statistical model and adaptive entropy-based segmentation. The automatic threshold segmentation algorithm detects defects rapidly using the threshold parameters obtained by modeling the image effectively, and the adaptive entropy-based segmentation algorithm effectively detects defects using ID information entropy. These two types of segmentation algorithms run in parallel, and their segmentation results are fused by an "and" operation. Thus, an improved image segmentation map with higher accuracy is obtained. Many experimental results and field applications show that the parallel combination segmentation algorithm is a stable and efficient segmentation algorithm, which improves the accuracy of the original segmentation algorithm that it contains.
Segmentation-based object categorization
Region growing
Range segmentation
Cite
Citations (8)
Image segmentation is important for target identification, watershed segmentation algorithm is widely used in image segmentation. In view of the over segmentation and sensitivity to noise problems of watershed segmentation algorithm, a image segmentation algorithm was proposed based on improved watershed algorithm. Firstly K-means clustering algorithm was used for the initial clustering segmentation, and then the watershed algorithm was used in image segmentation. The experimental results shows that the proposed algorithm has effect on solving the over-segmentation problem of watershed algorithm, And the image target is effectively segmented.
Segmentation-based object categorization
Region growing
Range segmentation
Cite
Citations (4)
Medical image plays an important role in the assist doctors in the diagnosis and treatment of diseases. For the medical image, the further analysis and diagnosis of the target area is based on image segmentation. There are many different kinds of image segmentation algorithms. In this paper, image segmentation algorithms are divided into classical image segmentation algorithms and segmentation methods combined with certain mathematical tools, including threshold segmentation methods, image segmentation algorithms based on the edge, image segmentation algorithms based on the region, image segmentation algorithms based on artificial neural network technology, image segmentation algorithms based on contour model and image segmentation algorithm based on statistical major segmentation algorithm and so on. Finally, the development trend of medical image segmentation algorithms is discussed.
Segmentation-based object categorization
Region growing
Range segmentation
Cite
Citations (10)
Image segmentation is a process, which divide an image into different regions, which are homogeneous in some characteristics. Image segmentation is the initial step in many image processing applications like Pattern recognition and image analysis. Image analysis includes object characterization and representation and feature measurement. Higher order task follows the classification of object. Hence, characterization, visualization of region of interest in any image, delineation plays a significant role in image segmentation. There are many segmentation algorithms available in the literature, which divide an image into number of regions based on some image features like pixel intensity value, color, texture, etc. These all algorithms are categorized based on the segmentation method used. They are Segmentation based on single or multiple thresh holding, Segmentation based on edge detection, Segmentation based on similar region and Segmentation based on clustering, Segmentation based on ANN and fuzzy logic technique etc. In this work, we have chose one algorithm from one segmentation category and implement the algorithm in MATLAB. The chosen algorithms are Otsu's algorithm, K-means, quad tree, Delta E, Region growing and fth algorithms. To check the performance of the algorithms, we have used 6 simple and complex images available in the literature. The obtained results shows the efficacy of the segmentation algorithms.
Segmentation-based object categorization
Region growing
Range segmentation
Connected-component labeling
Cite
Citations (27)
Image segmentation is an important image analysis technique, the segmentation result of which is critical to determine the performance of high-level modules in image processing system. After a brief introduction LLT image segmentation algorithm, focusing on analysis of the inadequacies of LLT algorithm, propose an improved ILLT image segmentation algorithm. Experiments show that the algorithm improves the efficiency of the image region segmentation.
Segmentation-based object categorization
Region growing
Range segmentation
Connected-component labeling
Cite
Citations (0)
An image segmentation algorithm based on ensemble learning is proposed.First,several different image segmentation algorithms are used to produce many intermediate segmentation results.Then the intermediate image segmentation results are integrated with ensemble learning.Finally,the integrated output is adopted to image segmentation.The thresholding segmentation method,region growing segmentation method and FCM segmentation method are used in the experiments.Experimental results show that the quality of the proposed image segmentation algorithm based on ensemble learning technology significantly outperforms the best individual member.And the proposed method can be a good solution to incomplete image segmentation problem.At the same time,the performance of the proposed method is often more robust than a single algorithm.
Segmentation-based object categorization
Region growing
Range segmentation
Ensemble Learning
Cite
Citations (1)