Image Recognition Application System Achievements
Technology to convert/edit images not only at the low-level unit of pixels, but also at the high-level units of strings and shapes
The combination solution of image recognition technology for image objects (characters, line drawings, shapes) and image processing technology

■Transparent Text PDF Conversion Using OCR

1. You can search for text strings in document images. The transparent text above the image will be inverted.
2. Detection of 90-degree, 180-degree, and 270-degree rotation of document images using text recognition.
3. Hybrid tilt detection by combining the tilt of line drawings and text rows.
4. Only the original image is displayed, so OCR recognition errors are not visible.
5. Transparent text conversion is possible for special documents such as official maps.

Super Low-Resolution Comic English Translation System

Overlay display of comics translated into English for QVGA size (240x320) mobile devices.
Note: The English translation program was created by a separate company.

Automatic Link System for Drawings and Text Information

Automatic generation of web pages linking drawing text information and explanatory text information
・Patent claims and drawings
・Drawings and title blocks
・Official cadastral maps and registration

■Intelligent Image Rescaling

System Using Character Recognition Library
Practicalization as a rescaling algorithm that changes the scaling factor for each target, as a comic rescaling tool for mobile phones

■Object Editing of Document Images

System Using Raster Editing Library
Edit document images (raster) on a per-object basis

■Image Recognition Applied Filters

Character image extraction filter, line width-based straight line recognition filter, photograph/graphic image area separation filter, heading extraction filter

■Automatic Binarization

Image Labeling Application System
By using a significantly faster image labeling algorithm compared to conventional methods, the image is binarized so that the number of cohesive regions (primitives or blobs) is optimized.
The number of regions changes based on the binarization threshold as follows (foreground strokes are considered to be low luminance = black):
The number of background regions is the opposite of the foreground regions.

  1. If the threshold is low and the image becomes blurred, the number of regions increases as lines are divided into dots.
  2. If the threshold is high and the image becomes smeared, the number of regions decreases as blank areas are filled in.
new binarization

■Document Image Object Conversion

Raster Editing Library Application System
An automated version of the object editing function described above.
Examples: (1) Automatic Beautification and Formatting of Line Drawings (converted into uniformly sized lines)
Examples: (2) Automatic Deletion of Meridians and Parallels (deletion that does not affect other shapes)

■Automatic Rotation

Character Recognition Library Application System
Detection of 90-degree, 180-degree, and 270-degree rotation in document images using text recognition.

■Simultaneous Acquisition of Multiple Documents by Digital Camera or Scanner

Character Recognition Library Utilization System
Automatically correct digital camera images to rectangular shapes, including those taken from diagonal angles.
Cut out multiple documents (assumed to be small like cards) on the screen, and automatically determine the correct orientation. Correct the tilt and acquire them as individual document images.

Before Digital Camera Correction Before Digital Camera Correction
After Digital Camera Correction After Digital Camera Correction

■Tilt Correction

Character Recognition Library Utilization System
Combination of tilt detection for line drawings and tilt detection for text lines using hybrid tilt detection.