Image Processing Reference
In-Depth Information
not very large, mostly because the classification module can still identify over-dilated images
or incomplete images.
5 Future research
In the future we would like to continue the research conducted so far. There are many areas
which can be improved and also, upon refactoring they can provide new functionalities:
• we intend to add a module in charge of collecting a user score for the result vector, which
can be used later for altering automatically the default weights in the majority voting mod-
ule. This way, the CBIR engine will be able to provide more representative results for the
• we have started to develop a user interface which will allow the user to configure the clas-
siication module according to his/her needs—this involves selecting the desired binariz-
ation/segmentation algorithms and tuning the majority voting modules. The user score
module and this module can be combined with a user management module so that the sys-
tem can recall the user's preferences;
• we also intend to insert a module that can analyze a certain image and compute how many
shadow areas it contains. This module would help in deciding which color space is more
adequate for each particular situation;
• the number of characteristics is very large; only the SIFT descriptors may go over 100,000
for 640 × 480 images. In order to be able to compute the results much faster, we are consid-
ering the possibility of adding a module in charge of reducing the dimensionality;
• in the document processing area, the system is currently cropping out the images from a
scan. In order to have more accurate results, we intend to add an OCR module, which can
extract the text content as well. The text can be later on reduced to keywords, which can be
used in the classification and retrieval process.
[1] Sebe, N. Feature extraction & content description—DELOS—MUSCLE Summer
School on Multimedia digital libraries, Machine learning and cross-modal techno-
logies for access and retrieval. . [Online] February 25, 2007. .
[2] Gevers T, Smeulders AW. Colour-based object recognition. Patern Recogn.
[3] Vassilieva, N. RuSSIR—Russian Summer School in Information Retrieval. [Online]
2012. htp:// .
[4] Cao X, Shen W, Yu LG, Wang YL, Yang JY, Zhang ZW. Illumination invariant ex-
traction for face recognition using neighboring wavelet coefficients. Patern Recogn.
[5] Bhavsar AV, Rajagopalan AN. Range map superresolution-inpainting, and recon-
struction from sparse data. Comput Vis Image Underst. 2012;116(4):572-591.
[6] Wikipedia. Local binary paterns. . [Online] November 08, 2011. .
[7] Lowe DG. Object recognition from local scale-invariant features. In: International Con-
ference on Computer Vision. 1999:1150-1157.
[8] Bay H, Tuytelaars T, Van Gool L. Speeded-up robust features (SURF). Comput Vis
Image Underst. 2008;110:346-359 Zurich, Leuven, Belgia: s.n.
Search WWH ::

Custom Search