The computed tomographic colonography (CTC) computer aided detection (CAD) program is a new method in development to detect colon polyps in virtual colonoscopy. While high sensitivity is consistently achieved, additional features are desired to increase specificity. The CTC CAD program we are developing outputs a 2D projection image for each detected polyp to mimic the radiologist reading. We designed a wavelet-based feature extraction method for the detection images in an attempt to filter out false positives. We computed a wavelet transform at levels 1-5, and extracted 150 features from the wavelet coefficients for each detection image. The feature vectors were then run through a feature and committee selection procedure with a support vector machine (SVM) classifier. We applied the proposed method to 44 patients, where the CTC CAD program identified 42 true detections from 28 polyps and 492 false positives. Three-fold cross-validation results showed that the proposed method could reduce the false positives by 69% while missing only one real polyp [96% sensitivity (27/28)]. If this technique were added to the filtering process of the CTC CAD polyp detector, the number of false positive detections could be reduced significantly.