A robust algorithm for text extraction in color video

We report on the development and implementation of a new algorithm for extracting text in digitized color video. The algorithm operates by locating potential text line segments from horizontal scan lines. Detected text line segments are expanded or combined with text line segments from adjacent scan lines to form larger text blocks, which are then subject to filtering and refinement. Text pixels within text blocks are then found by using bi-color clustering and connected-component analysis. A morphological contour smoothing, and a morphological resolution enhancement algorithms are then applied to the detected binary texts to enhance their visual quality. The implemented algorithm has fast execution time and is effective in detecting text in difficult cases, such as scenes with highly textured background, and scenes with small text. A variety of color images digitized from broadcast television are used to test the algorithm and excellent performance result was obtained.