| 摘要 |
A method of image recognition for detecting characters, is carried out in respect of at least one digital image captured by an image capturing device, and the method includes segmenting 205 the captured image so as to generate a segmented image. The segmented image comprises pixels which are selected in accordance with respective pixel attributes. The method further includes detecting, within the segmented image, a contiguous image region 215. The contiguous image region is an image region identified by those pixels adjacent to each other and corresponding to the same class of pixels identified in the segmenting. The method also includes modifying the detected contiguous image region to a predetermined line width so as to form a modified width image feature 225, generating a feature vector from the modified width image feature 230, classifying the generated feature vector in dependence upon a comparison between the generated feature vector and previously generated feature vectors that relate to predetermined characters, and detecting to which character the modified width image feature corresponds in dependence upon the classification of the generated feature vector 235. |
