Fig. 1. ROI, ROI central subarea, ROI subcentral subarea

Fig. 2. Variable standard deviation ROI center standard deviation-weighted Gaussian filter function regional distribution

Fig. 3. Comparison of results from different image definition calculation methods

Fig. 4. Data filling example diagrams. (a) In the case of 5×5 ROI segmentation, the standard deviation-weighted Gaussian filter function window (11×11) pixel filling; (b) 5×5 Sobel operator pixel filling example

Fig. 5. Gaussian filter calculation process

Fig. 6. 5×5 Sobel operator FPGA calculation process

Fig. 7. Sobel operator edge detection results of different sizes and directions. (a) Raw image; (b) 3×3 Sobel detection image; (c) 5×5 Sobel detection image; (d) four-direction 5×5 Sobel detection image

Fig. 8. Comparison of edge intensity of filtered images. (a)Raw noise image;(b) Gaussian filter image when σ=0.8; (c) Gaussian filter image when σ=3.2;(d) ROI subregion variable standard deviation Gaussian filter function image

Fig. 9. Experimental results comparison of no Gaussian filtering, traditional Gaussian filtering, and ROI subarea standard deviation-weighted Gaussian filtering. (a) The ROI size is 900×880; (b) the ROI size is 340×280; (c) the ROI size is 350×280