| 英文摘要 | Nowadays the amount of digital images and videos increases explosively with the development of high technology. And these multimedia documents contain a great deal of information which is valuable for many applications, such as information retrieval, image classification, data mining and etc. However, it is still very difficult for computers to understand the contents of images and videos. Among all the contents in images, text information has inspired great interests, since it can be easily understood by both human and computers, and leads to wide applications such as license plate reading, sign detection and translation, mobile text recognition, content-based Web image search, and so on. An integral Text Information Extraction system consists of three parts: text detection, text extraction and OCR. The first two parts, text detection and extraction, are critically important for the system performance. Text detection and extraction from images is a challenging problem due to the complexity of background, the variability of text position, size, font, color, polarity and line orientation. Based on this background, this dissertation presents an in-depth study on scene text detection and extraction by combining techniques in image processing and pattern recognition. Specifically, we propose a new text detection and extraction method, and the experimental results demonstrate the superiority of our methods compared with the state-of-the-art methods. The contributions of this dissertation are summarized below: Firstly, this dissertation proposes an interactive method to detect interesting texts in natural scene images. We first draw a line to label a region which contains the texts we want to detect. Then a coarse-to-fine strategy is adopted to detect the texts in this label region. For coarse detection algorithm, we use as concise as possible to ensure high speed and recall rate. For fine detection, we focus on improving the accuracy of the detection results. Overall algorithm to ensure both a high recall rate and accuracy, and ensure the high speed of algorithm. Experimental results demonstrate very promising performance on detecting texts in complex natural scenes. Secondly, this dissertation proposes a text extraction algorithm incorporating local information. We regard text extraction as segmenting the image and removing noises, and then a robust text extraction menthod incorporating local information is proposed. First, we get the gray image from the o... |
修改评论