作者: Satoshi Naoi , Yuan He , Siyuan Chen , Jun Sun
DOI:
关键词: Digital image processing 、 Information retrieval 、 Structured document 、 Image texture 、 Artificial intelligence 、 Feature detection (computer vision) 、 Image processing 、 Contextual image classification 、 Template matching 、 Feature extraction 、 Standard test image 、 Digital image 、 Image registration 、 Pattern recognition 、 Feature (computer vision) 、 Automatic image annotation 、 Computer science 、 Binary image
摘要: Following the recent trend in using low level image features classifying document images, this paper we present a novel approach for structured classification by matching salient feature points between query and reference images. Our method is robust to diverse training data size, formats qualities. Through points, registration available as well. Although aimed large domain of our already achieved zero error rates tests on benchmark NIST tax form databases.