Fig. 1. Algorithm flowchart

Fig. 2. Schematic of three-layer window structures. (a) Small target;(b) around noise;(c) large target

Fig. 3. Experimental results. (a) Original grayscale; (b) threshold too small

Fig. 4. Experimental results. (a) Threshold too large; (b) Kittler algorithm

Fig. 5. Detection results of proposed method under different voltages. (a) 1 V; (b) 2 V; (c) 3 V; (d) 4 V

Fig. 6. Standard deviation results of target centroid coordinates for corona detection using proposed method and morphological method

Fig. 7. Accuracy results of target centroid detection for corona detection using proposed method and morphological method

Fig. 8. Multi-point corona detection results of proposed method under different voltages. (a) 1 V; (b) 2 V; (c) 3 V; (d) 4 V

Fig. 9. Standard deviation results of multi-target centroid coordinates for corona detection using proposed method and morphological method

Fig. 10. Precision results of multi-target centroid detection for corona detection using proposed method and morphological method

Fig. 11. Visual contrast results of different methods. (a) Original image; (b) low-pass filtering; (c) Facet filtering; (d) Facet filtering and local contrast enhancement

Fig. 12. Denoised images. (a) Denoised image local contrast; (b) denoised image gradient resolution

Fig. 13. Processing results of deep learning methods. (a) Training data; (b) denoising result image