Study of Content Based Image Retrieval Systems

Jitendar, Rupavath (2016) Study of Content Based Image Retrieval Systems. MTech thesis.

[img]PDF (Full text is restricted upto 05.04.2020)
Restricted to Repository staff only



In recent years, very gigantic collections of pictures and movies have grown rapidly. In parallel with this development content based recovery and querying the indexed collections are required to access visual knowledge. The growing demands for snapshot picture recovery in multimedia subject akin to crime prevention, wellbeing informatics and biometrics has pushed utility builders to go looking ways to control and retrieve images more successfully. Retrieving pictures situated on content material corresponding to texture and colour remains to be a challenging issue. Texture and colour are the intangible knowledge embedded in picture. Representing this understanding’s properly is significant so as to obtain better retrieval results. Additionally, the latest wavelet headquartered photo compression science has been obvious as a new solution to store thousands of pictorial knowledge within the restrained space of the hardware capabilities. Wavelets were verified to be superior in terms of compression compared to earlier compression ways.
In this thesis, I executed a method to retrieve wavelet based compress pictures situated on its colour and texture features. Colour and texture retrieval was once performed utilising exceptional classifiers like Euclidean Distance, Manhattan Distance and Standard Euclidean Distance respectively. Texture classification is very predominant in photograph evaluation. Content-established recovery, inspection of surfaces, object recognition by texture, document segmentations are few examples the place texture classification performs a fundamental role. We evaluate the effectiveness by making use of Precision and take into account assessments and retrieval time on quite a lot of units of experiment pics. Texture retrieval performed on outcome from colour matching provided higher precision and consider ratings compared to texture retrieval carried out ordinarily compressed picture within the database. Two of the principal components of the visible understanding are texture and colour. In this thesis, content-founded restoration is provided that computes texture and colour similarity among snap shots. The foremost manner is headquartered on the difference of a statistical strategy to texture analysis. An top-quality set of five 2d-order texture facts are extracted from the Spatial grey Stage Dependency Matrix of each snapshot, in an effort to render the function vector for each picture maximally informative, and yet to receive a low vector dimensionality for efficiency in computation

Item Type:Thesis (MTech)
Uncontrolled Keywords:Content-based recovery; Grey Level Dependency Matrix; Similarity matching; Texture classification; Feature vector; Wavelet; Precision and Recall
Subjects:Engineering and Technology > Electronics and Communication Engineering > Image Processing
Divisions: Engineering and Technology > Department of Electronics and Communication Engineering
ID Code:9281
Deposited By:Mr. Sanat Kumar Behera
Deposited On:06 Apr 2018 16:11
Last Modified:06 Apr 2018 16:11
Supervisor(s):Meher, Sukadev

Repository Staff Only: item control page