Extraction and instant translation of text from static images and dynamic videos is a very challengeable young research area, which nevertheless attracts increasing researchers' attentions. The traditional scanner-based document analysis techniques provide a good reference and technical background, but they cannot be adapted directly on camera- captured images or videos. It is found that the text extraction from camera-captured images or video frames had highly suffered from low resolution, blur, perspective distortion, and various brightness or grayscale, as well as complex layout and interaction of the content with background, which had become the inevitable obstacles in realizing the visual translation. Towards solving the aforementioned bottleneck problems, a novel technique of detecting and extracting text regions on camera-captured and video images for realizing more accurate text recognition is proposed. As a result, the higher accuracy of camera-based visual translation is witnessed from the obtained testing results. ©2008 The Institution of Engineering and Technology.