Paper Title:
Algorithms for Manipulating Compressed Images
Authors:
Brian C. Smith and Lawrence A. Rowe
Main Point(s):
The paper proposes a new technique that implements operations directly
on compressed JPEG data, resulting much faster performance than the
brute-force way of decompressing first, processing and compressing
the final result. Basic algebraic image operations of pixel-wise and scalar
addition and multiplication are shown and used to implement two common
video transformations: dissolve and subtitle.
Possible Use(s):
Since the operations on the compressed image domain is much faster, we
can use the schemes to do real-time digital video data.
Extensions:
More image operations, such as geometric transformations and filtering,
and image understanding and computer vision algorithms can be explored
in compressed image domain.
Jing Huang's CS631 course page