Super-Resolution Reconstruction Technique And Its Application In Defect Inspection Of Flip Chip

Flip chip technology is used extensively in microelectronics packaging.Since the integration density of the chip is constantly increasing.The defect detection and failure analysis of the chip will become challenge.In this thesis,SAM technique is used to inspect the solder bumps which were placed between the die and carrier.However,the resolution of the acquired SAM image is often insufficient for the recognition of the defective bumps due to the influence of the noise,degradation and the limitation of the imaging lens.Thus,we will foucus on the resolution reconstruction procedure of the SAM images based on the SR(Super Resolution)technique.Following is the main content:A cascade of images with different scales is contructed to produce the highresolution/low-resoluton patch pair according to the phenomenon of patch redundancy in a single image,a high resolution SAM image is then acquired by applying the proposed SR technique on the original image,which has combined the exampled-based SR technique and the classical SR technique.The binary stencil based on the image gradient is adopted to segmenment the bumps from the background.Several stasticial values of the relevant area is calculated,and the improved K-means clustering algorithm is applied to realize the recognition of the defective bumps.Another SR technique is studied for the resolution-enhancement of the SAM image.The decomposition wavelet is adopted to introduce an intermediate stage to enhance the high frequency information of the original SAM image.And all the frequency subbands is combined together by the fusion wavelet to produce the high SAM image.Every solder bump area is extracted by the template match algorithm and the PCA(Principle Component Analysis)is applied to recombine the original abundant features.Finally,the diagnosis of the defective bumps is realized by using the SVM classification algorithm.In this thesis,the super-resolution technique is used to reconstruct the highresolution SAM image,and provide a reliable diagnoise system for the regocnition of the bumps.The results show that the reconstructed SAM image containing much more feature information of the bumps,and the accuracy of the system's ability to recognize the bumps is increased.