Patent №
US 5,535,013
Granted
—
Owner
—
Lab
—
AI components
1
vision
Assignment
None on record
Dataset
AIPD
2023_r1 edition
Application
08308517
An image signal compression apparatus which divides image data into plural blocks, encodes each block, orthogonally transforms the image data by each block, and outputs the transformation coefficients. Blocks which contain text and blocks which contain dot images are discriminated from the features of the absolute values of the AC coefficients, which are a subset of the transformation coefficients. One of plural quantization tables is selected for each block based on the discrimination result, and by adaptively quantizing the transformation coefficients using the selected quantization table, text images can be reproduced free of fuzziness, and good expanded images of the dot areas can be obtained with high compression by reducing the high frequency component. When the predetermined high frequency transformation coefficients, which are a subset of the above transformation coefficients, are greater than the predetermined low frequency transformation coefficients and a predefined threshold value, the block is discriminated to include a dot image area. When the predetermined low frequency transformation coefficients, which are a subset of the above transformation coefficients, are greater than a predefined threshold value, the block is discriminated to include a text image area.