&lt;title&gt;Field-test results of an image retrieval system for semiconductor yield learning&lt;/title&gt;

Karnowski, Thomas P.; Tobin, Kenneth W.; Arrowood, Lloyd; Ferrell, Regina; Goddard, Jim; Lakhani, Fred

doi:10.1117/12.429362

Cited by 3 publications

(1 citation statement)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…One way to determine the source would be to search the historical defect image database to find previously diagnosed cases in which the defect 10 resembles the new one. Before this database can be searched for defects with similar appearance using a technique such as content-based image retrieval [16], the defects within the database must first be redetected. Finding defects without the aid of a reference image is relatively easy for a human but is an extremely challenging task for a computer vision system.…”

Section: The Inspection Taskmentioning

confidence: 99%

Defect pattern detection using a new rule-based approach

Shankar¹

View full text Add to dashboard Cite

Automated inspection of semiconductor defect data has become increasingly important over the past several years as a means of quickly understanding and controlling contamination sources and process faults, which impact product yield. To address the issue of too much data and too little time, automation technologies in defect detection and review are being developed by universities, laboratories, industry, and semiconductor equipment suppliers. In this thesis, a new rule-based approach is proposed to segment defect images. Several segmentation techniques already exist but they often focus on the constraints of a specific application and therefore they lack of generality and flexibility. This limits the use of computer vision in all those tasks where the visual data content and the purpose of the defect analysis are not known a priori. Moreover, the limited generality increases the costs for the design of unsupervised defect image analysis systems. To overcome these limitations, it is proposed to decompose a general defect image segmentation problem in four levels. The lowest levels have a high degree of generality and are inspired by the perceptual mechanisms of human vision. Their main role is to simplify the input data and to extract the perceptually meaningful information. The proposed process is sufficiently general to be applied to a wide class of applications and input data without the need of human supervision. In order to achieve a higher degree of autonomy, generality and a more flexible solution, the proposed defect image segmentation approach adopts a hierarchical perspective. The visual information in a frame is viewed as being composed of patterns, referred to as objects, which are built from simpler sub-patterns, referred to as regions, which are themselves built from yet simpler primitives corresponding to the pixels in the defect image. vi

show abstract

Section: The Inspection Taskmentioning

confidence: 99%

Defect pattern detection using a new rule-based approach

Shankar¹

View full text Add to dashboard Cite

show abstract

<title>Detection of semiconductor defects using a novel fractal encoding algorithm</title>

et al. 2002

Self Cite

View full text Add to dashboard Cite

This paper introduces a new non-referential defect detection (NRDD) algorithm for application to digital images of semiconductor wafers acquired during the manufacturing process. This new algorithm is composed of two major steps: (1) defect detection via the use of fractal image encoding and (2) enhanced defect boundary delineation using active contours. One primary application for this technology is the redetection of defects within archived databases of historical defect imagery. Defect images are commonly stored by semiconductor manufacturers for future diagnostic purposes, but reference (golden) images are usually unavailable. The ability to automatically redetect a defect is crucial in an automated diagnostic system that uses the historical defect images for defect sourcing. Results are presented for four large data sets of semiconductor images. Three of these data sets are composed of scanning-electron microscope (SEM) images and the fourth contains optical microscope images. Performance criteria were created that score the NRDD segmentation result as a percentage based on a comparison to a manually outlined version of the defect. The overall NRDD score across all four databases ranged from 50% to 84% using the same set of manually-determined parameters on all images within each database. By using an automated parameter setting algorithm these performance values improved to 57% to 92%. The NRDD algorithm performance depends, in part, on the size of the defect and the level of complexity of the background of the semiconductor image.

show abstract