If I have two techniques which handle text detection from image, on depend what can I compare of the methods?
Hi Abdelrahiem,
Briefly, once the features get extracted from the image, perform data preprocssing stage, then compare between your methods under the evaluation of cross-validation. Based on the cross-validation results (ROC, F-measure, etc.) you can build a clear understanding about performance of the utilized methods.
HTH.
Samer
I agree wih Samer Sarsam .... the proposed methodology is very robust.
Having ground truth image (data) for any database would be very useful to evaluate any algorithm's strength. If not then you can do it manually though tedious and time consuming and some time error-prone. For evaluating the effectiveness of your procedure, like others say, you can use Precision, Recall, F-measure etc.