Abstract:
This thesis is aimed to propose the image contrast enhancement method based on the decomposition
of image into two layers i.e, at texture and structure layer. Contrast enhancement
being the widely used tool in computer vision can be applied either locally such as tone
curve adjustment or globally such as dehazing. Current techniques fails to enhance the compressed
input images as compression artifacts also become visible as the original images are
enhanced at the same time. This dissertation is focusing on solution of this problem firstly
by discussing the current state-of-the-art techniques used for artifacts removal on contrast
enhanced images. Initial work shows that much more efforts are needed in this direction.
The major issue is the visibility of invisible compression artifacts in low-bit images along
with the other image content enhancements. We confronted here two non-proportional tasks
concurrently i.e., developing a well-sighted image by enhancing its contrast while compressing
its compression artifacts. Data processing in sequential manner e.g pre-processing and
post-processing technique is highly unlikely as during pre-processing loss of valuable content
of the image with low-bit rates will occur while post-processing will cause damage to
high-bit rate content. So it is near to be impossible to achieve best formulation because of
addressing both image content and noise simultaneously.
We have proposed a methodology to deal with this issue, that is to decompose image into
structure and texture layer and then apply contrast enhancement techniques of these layer
separately. The proposed algorithm suppress the artifacts appearing in the JPEG images
separately i-e structure component and texture component both are processed independently
and the compression of artifacts in parallel with contrast enhancement, that is inadvertently
visible when contrast of JPEG image is enhanced. The results clearly demonstrate a novel
improved approach to increase the contrast of image by avoiding the distortion and producing
better results comparable to the generic deblocking techniques.
