OCR-oriented Master Object for Text Image Captioning.
Wenliang TangZhenzhen HuZijie SongRichang HongPublished in: ICMR (2022)
Keyphrases
- image data
- scanned documents
- region of interest
- printed documents
- image analysis
- keypoints
- single image
- bounding box
- image content
- image regions
- post processing
- image classification
- target object
- pixel level
- image features
- information retrieval
- multiple objects
- scanned images
- text recognition
- input image
- image segments
- hough transform
- test images
- similar objects
- lighting conditions
- partial occlusion
- d objects
- object model
- image representation
- edge detection
- visual appearance
- printed text
- foreground and background
- optical character recognition
- character recognition
- spatial relationships
- image collections
- feature points
- high resolution
- complex scenes
- spatial information
- back projection
- segmentation algorithm
- document processing
- low level
- image retrieval
- keywords
- normalized correlation
- multiscale