Denoising method for micro-focus X-ray images for integrated circuit packaging inspection

Abstract

The denoising method of micro-focus X-ray images for integrated circuit packaging detection disclosed by the invention includes the following steps: Gaussian preprocessing of the input image; using a differential curvature map to divide the micro-focus X-ray observation image into flat areas and Detail area; use the partition results and differential curvature to calculate the regularization operator and regularization parameters; construct the objective function, and use the approximation-based variable step algorithm to quickly iteratively solve the objective function. The denoising method of the present invention is solved by a fast iterative solution algorithm with variable step size based on approximation. The algorithm is fast and meets actual industrial requirements.

Description

technical field [0001] The invention relates to the field of image denoising, in particular to a denoising method for micro-focus X-ray images oriented to integrated circuit package inspection. Background technique [0002] The focus of the micro-focus X-ray machine is only 5-10 μm, which can realize the detection of tiny and fine structures, and is often used for component detection and defect detection in integrated circuit packaging. Since the detection objects in integrated circuit packaging are often a variety of different metals and their combinations, different metals have different absorption conditions for X-rays, so micro-focus X-ray images can non-destructively detect the internal structure of circuits, but also due to various metals The absorption of the micro-focus X-ray image is relatively close, which makes the micro-focus X-ray image show the characteristics of low contrast. At the same time, due to the influence of noise, it is difficult to accurately detect...

AI Technical Summary

Problems solved by technology

The micro-focus characteristics of micro-focus X-ray machines make the interaction of noise more complicated than the existing common X-ray image noise models; at the same time, for micro-focus X-ray images for integrated circuit detection, due to subsequent needs Accurate detection of integrated circuit parts and defects, so the requirements for image detail retention are high while effective denoising, and its image has the characteristics of many weak details, low contrast, and low signal-to-noise ratio, which increases denoising difficulty

[0004] Most of the existing research only focuses on single Gaussian noise or Poisson noise, which is not applicable to micro-focus X-ray images for integrated circuit packaging inspection. Single Gaussian noise or single Poisson noise is much more complicated; in terms of regularization methods, most of the existing research focuses on the design of a single adaptive scheme of regularization operators or regularization parameters, which cannot be well combined. It takes a lot of time to determine the specific parameters, and the balance between noise removal and detail preservation cannot be guaranteed, and the requirements for denoising in the actual industrial integrated circuit packaging inspection cannot be well met.

Images