Abstract: With the advances of content-based retrieval technology, people have realized the importance of semantics for text embedded in color image and video. So, it needs tools to effectively extract the color image and video text. The paper presents a new color edge detector ColorEdge which operates on the overall effect of three channels of RGB color space, and a new color image and video text extraction approach EDA which combines color edge detector with area-based image segmentation technology is worked out. It shows that the text edges detected by ColorEdge are clear, integrated and being easily separated from the background later, while EDA has high extraction ratio and is fairly robust to the size and color of text characters.