Image Processing Reference
ted is insignificant and will likely be fixed through simple padding in subsequent processing.
The image on the right, on the other hand, was classified as a partial true positive. While the
localized area does contain most of the NL, some essential text in the left part of the NL is ex-
cluded, which will likely cause failure in subsequent processing.
FIGURE 13 Complete (left) vs. partial (right) true positives.
In Figure 14 , the left image technically does not include the entire NL, because the list of in-
gredients is only partially included. However, we classified it as a complete true positive since
it includes the entire table on nutrition facts. The right image of Figure 14 , on the other hand,
is classified as a partial true positive, because some parts of the nutrition facts table is not in-
cluded in the detected area.
FIGURE 14 Complete (left) vs. partial (right) true positives.
Of the 266 images that contained NLs, 83 were classified as complete true positives and 27
were classified as partial true positives, which gives a total true positive rate of 42% and a false
negative rate of 58%. All test images with no NLs were classified as true negatives. The re-
mainder of our analysis was done via precision, recall, and specificity, and accuracy. Precision
is the percentage of complete true positive matches out of all true positive matches. Recall is
the percentage of true positive matches out of all possible positive matches. Specificity is the
percentage of true negative matches out of all possible negative matches. Accuracy is the per-
centage of true matches out of all possible matches.
Table 1 gives the NL localization results where PR stands for “precision,” TR for “total re-
call,” CR for “complete recall,” PR for “partial recall,” SP for “specificity,” and ACC for “ac-
curacy.” While total and complete recall numbers are somewhat low, this is a necessary trade-
of of maximizing specificity. Recall from Section 1 that we have designed our algorithm to
maximize specificity. In other words, the algorithm is less unlikely to detect NLs in images
where no NLs are present than in images where they are present. As we argued above, lower