Improving OCR-Based Image Captioning by Incorporating Geometrical Relationship.
Jing WangJinhui TangMingkun YangXiang BaiJiebo LuoPublished in: CVPR (2021)
Keyphrases
- input image
- multiscale
- image features
- image retrieval
- image classification
- image content
- single image
- image segmentation
- image data
- scanned documents
- image analysis
- template matching
- lighting conditions
- spatial information
- post processing
- feature points
- image pixels
- segmentation method
- test images
- error correction
- keypoints
- grey level
- vector field
- region of interest
- optical character recognition
- character recognition
- pixel values
- image structure
- gray scale images
- scanned images
- document images
- hough transform
- image regions
- gray scale
- image representation
- low level
- high resolution
- preprocessing
- object recognition
- video sequences