The Study Of Segmentation And Recognition For Video Image Character

The mankind recognizes various kinds of characters and images with eyes and brain, and can get rid of various kinds of interference and distortion. So, it has important theory meaning and practical value for recognition and intelligent pattern-recognition to study the course of machine's humanoid recognition.. Based on the structure and stroke characteristic of Chinese character, this thesis used wavelet transformation and mathematical morphology to segment video image character. Humanoid recognition try hard to make the computer imitate the human vision based on the characteristic of the machine recognition. The primary contents of this thesis are:1) Image segmentation pretreatment. It uses wavelet transformation and morphology to segment video image character. The character usually has certain direction information. Two-dimension wavelet can draw the edge detail in every direction. Morphology can segment the character area.2) Code scheme for Chinese character. Ten-code scheme for humanoid recognition is presented. After studying human's code scheme, Wubi code and Sijiao code and Taiji code, based on the characteristic of the machine recognition, this thesis presents the ten-code scheme for humanoid recognition.3) Recognition algorithm. On the basis of resolving to Chinese character stroke, this thesis presents the abstraction algorithm which recognition code used according to the character characteristic.The research of this text shows: Segmentation for video image character can draw characters of image effectively, especially individual character. The slope characters can be segmented too. The recognition code scheme can code and distinguish the characters very well according to recognition code that the characteristic of the character is drawn.Summarize the advantage and deficiency of humanoid recognition code scheme of Chinese characters in the thesis. Proposed some thinking and prospect in studying.