During the development of a new printer product, units of the product undergo extensive testing, during which thousands of print quality test pages are printed. Compared with visual evaluation of these pages by print quality experts, it is more efficient and cost-effective to scan the pages, and then perform image analysis to assess their quality. In this project, we develop an algorithm for the detection of a specific type of print quality artifact: local defects, and the prediction of the overall print quality that would be assigned by an expert observer to prints that exhibit such defects. Local defects are print defects in the form of spots and patches. In the detection procedure, the local standard deviation of the scanned pages is firstly computed to find the candidate regions where local defects are likely to occur. Subsequently, an automatic thresholding algorithm, namely valley-emphasis thresholding, is applied to find the local defects. Multiple features of each local defect are further calculated. With a database of print samples for which grades have been assigned by a print quality expert, a print quality predictor is trained by the support vector machine (SVM) method. For a new print sample, the detection procedure firstly finds the local defects and their features, and then the print quality is determined by the trained SVM model.