In this paper a new text extraction algorithm isproposed. In real scenes the text is usually overlapped or is partof the background. To identify the text regions, in complexconditions, a method exploiting a “multi-resolution feature basedmethod” for extracting text with undefined dimension has beendeveloped. Once identified, the multi-resolution information aremerged and skimmed through a set of Support Vector Machines(SVM). The tests and the comparisons with other techniques,performed on heterogeneous images, have shown theeffectiveness of the proposed.
|Numero di pagine||2|
|Stato di pubblicazione||Published - 2009|
All Science Journal Classification (ASJC) codes