Image processing method in bump height measurement in chip packaging

Abstract

Bumps are important structures used to connect chiplets to the interposer or package substrate in advanced chip packaging. Bump-height inconsistency will lead to problems in chip interconnection. Thus bump height detection plays a significant role in chip package quality inspection. The optical triangulation measurement system is preferably used to measure bump height in industry due to its high accuracy and high efficiency which meets the need of on-line inspection. In the system, the speed of image processing is one of the key factors which influence the measurement efficiency. Traditional means of bump height measurement often sacrifice speed to achieve high-precision measurement which leads to low efficiency. In this paper, a fast optical stripe feature extraction method is proposed which combines the gray centroid method with the Steger method in laser strip image processing in the optical triangulation system. This method can effectively reduces the number of iterations in image processing so as to accelerate the measurement speed. Firstly, the method uses the grey centroid method to calculate the center point coordinates of the light strip. Secondly, the coordinates derived are used as the initial point selection range for Steger method. Finally, the center points of the laser strip are obtained using the Steer method. The experimental results show that the proposed method reduces the time to extract the centroid of light strip from 5256 ms to 3184 ms under the same hardware conditions without compromising the accuracy compared to the conventional Steger method. To demonstrate the effectiveness in practical application, the centroid for 10 stripes was measured using the proposed method. Experimental results show that the proposed method significantly reduces the extraction time for centroid coordinates by about 40% without notable accuracy degradation.