Automatic knowledge extraction from images

Technology is constantly evolving at a fast pace, especially for Internet Industries where it has already impacted every developed country. Thus, the search engine of the Web needs to improve and bring it to a whole new level. The accuracy and performance of the search engine plays a big role in ens...

Full description

Saved in:
Bibliographic Details
Main Author: Wee, Jia Xin
Other Authors: Mao Kezhi
Format: Final Year Project
Language:English
Published: 2018
Subjects:
Online Access:http://hdl.handle.net/10356/75353
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Nanyang Technological University
Language: English
Description
Summary:Technology is constantly evolving at a fast pace, especially for Internet Industries where it has already impacted every developed country. Thus, the search engine of the Web needs to improve and bring it to a whole new level. The accuracy and performance of the search engine plays a big role in ensuring reliable data presented to the user. The need to discover new innovations and development is vital to bring convenience to users. The problem with traditional methods of indexing images led to the rise of interest in retrieving images with automatically derived features. Professional groups like journalists, design engineers and art historians rely heavily in retrieving a collection of images from a variety of domains [1]. We will adopt the search technology of Content Based Image Retrieval to represent high level semantic content depicted in images. This technique can identify or retrieve a collection of labelled images to bring the best level semantic annotations [2]. While different users will have different requirements for their usage, it will be useful to categorize image queries of abstraction into three levels: primitive, logical and abstract features of the image scenes depicted [1]. In this report, a detailed and refined search will be presented to display how the process of retrieving desired image from a large collection based on features can be extracted automatically. It will be programmed in a way to think like a human with the machine learning theories and image processing principles. We will combine the steps on the process of extraction for the machine to operate like human. It involves the extraction of data analysis from the image input and the text that is tagged to the image found. They are usually short descriptions or titles attached to the desired image. An ideal and effective outcome will come from the verification results with the program which maximizes Content Based Image Retrieval (CBIR) capabilities.