Research On Character-Localization In Images

With the development of the multimedia, the application of digital image (video) in all the fields is more and mere. Now the ability of finding the needed information from the many images (video) is required by the people. For the image, the information of the characters in it reflects part important content of this image. Automatic Image index based on the characters with the semantic character in the image is one of the important research fields of the image index technologies.Reference [12] presents that the texture of the edges of the character has the directions, that is horizontal direction, vertical direction, and slant directions. With the distinguishing characteristics of character' s texture (such as horizontal lines, vertical lines, or slant lines in a character) that can be extracted directly, the character-regions are segmented from their background quickly, and the image-noises rising during the processing period can be removed by morphological filter. With this method, the compressed bit-streams, which are encoded by DCT-Based encoding algorithm, can be processed directly to locate the character-regions in the images, just a very small amount of decoding is required So, the amount of data which want to process is smaller, the processing speed is faster and the demand of computer memory is less. Now the new energy definition combined with the nearby blocks to instead that in reference 12 and the adaptive dynamic threshold instead of the fixed threshold are presented. Experiments show that the amended method is better than that in reference 12. As the same time the method can combine with the wavelet transformation to locate the characters. By multi-resolutionanalysis and pyramid decomposition, the edge components with different spatial resolutions and different directions can be acquired, among which, the detail components have the most distinguished texture features standing for the object region, Then by further morphological operations, the useless information is greatly decreased and the last object text region is acquired.Experiments show that he correct-localizations rate of this model applied in the dot-domain and wavelet-transformation-domain is higher.