Research on Image Smoothing Diffusion Model With Gradient and Curvature Features

Abstract

In this paper, two image smoothing models are proposed for the visual inspection of high-density flexible IC package substrates with strict requirements on line width and line distance which are applied to the de-noising of high-density flexible IC package substrate images. First of all, the two models proposed in this paper combines the level set curvature feature of the image with gradient threshold, using more abundant second-order differential information as the detection factor to remove the noise in the image. Second, the theoretical analysis shows that the de-noised image obtained by the two models proposed can retain more detailed texture information and edge information of the original image. What is more, the experimental analysis shows that the proposed models have the highest structural similarity and peak signal-to-noise ratio, and have a relatively high edge-preserving index and the lowest mean squared error compared with other models. In particular, the de-noised image through Model 1 has the highest structural similarity and peak signal-to-noise ratio, as well as the lowest mean squared error. The de-noised image through Model 2 has a relatively high edge retention index. The methods proposed in this paper can effectively remove the noise of the image of the high-density flexible IC package substrate and can retain the original details and edges information of the image.