Automatic text detection in complex color image

Detection of text in color images of complex colored background is a very challenging problem. In this paper, an efficient automatic text detection method-fusing multi-feature Is proposed. First, we generate candidate text regions by merging bounding blocks, which are extracted using color feature in spatial color quantized map and edge feature in edge map gotten by sobel operators followed by mathematical morphology operators. The corner feature and heuristics based edge and color features are then used to eliminate the false candidates. Experimental results on book covers and natural scene Images show that this method can detect text including Chinese and, English characters accurately. In addition, this method is expected to be more accurate for Chinese character detection, because the verification uses more prior knowledge aimed at Chinese characters.