The present invention provides a method for detecting and classifying
block edges from DCT-compressed images, and more particularly, a fast
method for classifying the edge direction of each block in the
DCT-compressed images. The method includes: extracting DCT coefficients
by blocks constituting the compressed image; and applying an arithmetic
operation defined for each direction component to the DCT coefficients,
and comparing the results of the arithmetic operations to determine the
edge direction component. The present invention is directly applied in
the DCT domain based on the quantitative analysis on the contribution
factor of each DCT coefficient to formation of the directional pattern of
the block edge, thereby drastically reducing the computational complexity
of arithmetic operations necessary to the detection and classification of
block edges.