COME: Clip-OCR and Master ObjEct for text image captioning.
Gang LvYining SunFudong NianMaofei ZhuWenliang TangZhenzhen HuPublished in: Image Vis. Comput. (2023)
Keyphrases
- image data
- image features
- scanned documents
- image regions
- multiscale
- image analysis
- complex scenes
- multiple objects
- single image
- input image
- normalized correlation
- printed documents
- lighting conditions
- feature points
- image classification
- post processing
- bounding box
- segmentation method
- image retrieval
- visual appearance
- target object
- region of interest
- pixel level
- image collections
- error correction
- high resolution
- spatial relations
- low level
- web images
- image content
- segmentation algorithm
- image representation
- spatial relationships
- test images
- keypoints
- image segmentation
- color images
- scanned images
- inverse halftoning
- text recognition
- moving objects
- document processing
- text information
- edge detection
- document analysis
- foreground and background
- optical character recognition
- complex background
- object models