Summary of PhD thesis Computer science: Improve the efficiency of contentbased image retrieval through weight adjustment technique of the distance function

The thesis proposes a number of image retrieval methods to improve image retrieval accuracy. These methods will address issues of reducing semantic gaps between low-level features and high-level concepts of images.
Nội dung trích xuất từ tài liệu:
Summary of PhD thesis Computer science: Improve the efficiency of contentbased image retrieval through weight adjustment technique of the distance function MINISTRY OF EDUCATION VIETNAM ACADEMY OF SCIENCE AND TRAINING AND TECHNOLOGY GRADUATE UNIVERSITY OF SCIENCE AND TECHNOLOGY ……..….***………… DAO THI THUY QUYNH IMPROVE THE EFFICIENCY OFCONTENT-BASED IMAGE RETRIEVAL THROUGH WEIGHT ADJUSTMENT TECHNIQUE OF THE DISTANCE FUNCTION Major: Computer Science Code: 9 48 01 01 SUMMARY OF PHD THESIS Ha Noi – 2019 1The thesis has been completed:Graduate University Of Science And Technology - Vietnam Academy OfScience And TechnologySupervisor 1. Assoc. Prof. Dr. Ngo Quoc TaoSupervisor 2. Assoc. Prof. Dr. Nguyen Huu QuynhReview 1:Review 2:Review 3:The thesis will be defended at the Board of Examiners of GraduateUniversity Of Science And Technology - Vietnam Academy Of ScienceAnd Technology, at……………….on………………………….The thesis can be explored at:- Library of Graduate University Of Science And Technology- National Library of Vietnam 2 PREFACE1. Motivation of the thesis Image database is becoming more and more popular in various applicationfields such as remote sensing, crime prevention, medicine, education ... Theevolution of image acquisition, transmission and storage techniques has allowedthe construction of very large image databases. These factors have prompted theattention of the image retrieval community to propose ways to effectively exploitthis image database. Content-based image retrieval techniques are time-consuming, costly, anddepend on the subjective perception of the descriptors. Moreover, when changesare needed, this system is difficult to implement. To overcome this problem, in the early 1990s, content-based image retrieval(CBIR) was launched. The basic idea of this approach is to use automated visualfeature extraction techniques to produce content descriptions from images, suchas color, texture, and shape characteristics. These content descriptions are usedfor indexing images. There are many content based image retrieval systems thathave been proposed. However, many experiments on CBIR systems indicate thatlow-level content often fails to describe high-level semantic concepts that appearin the users mind. The performance of CBIR systems is still far from the usersexpectations. Therefore, the thesis chooses the topic of Improve the efficiency of content-based image retrieval through weight adjustment technique of the distance function tocontribute to solving existing problems.2. The objective of the thesis The thesis proposes a number of image retrieval methods to improve imageretrieval accuracy. These methods will address issues of reducing semantic gapsbetween low-level features and high-level concepts of images.3. The major contributions of the thesis Propose SRIR method (Semantic–Related Image Retrieval method),AWEIGHT method (An efficient image retrieval method using adaptiveweights).4. Thesis organization The contents of the thesis are presented in three chapters. Chapter 1 introduces the fundamentals of content-based Image retrieval. Chapter 2 presents the semantic-related image retrieval method. Chapter 3 presents an efficient image retrieval method using adaptiveweights. Final, the thesis presents conclusions and future research. 1 Chapter 1. OVERVIEW OF CONTENT-BASED IMAGE RETRIEVAL1.1. Introduction Various types of multimedia sources are increasing rapidly, such as visual datain smartphones, 2D / 3D applications, web content, etc. Therefore, the need forimage applications is more important than ever. However, visual media requires asignificant amount of processing and storage. It requires efficient methods toindex, store, analyzes and retrieval visual information from image databases. Therefore, retrieval images to ensure quickness, accuracy, and efficiencybecomes one of the challenging tasks.1.1.1. Text-based image retrieval The original approach for image retrieval is based on the text that describes theimage, in which the images are indexed by keywords, topics or classificationcodes. These keywords, topics or classification codes are used in the retrievalprocess. However, with large image databases, the difficulties faced by the text-based retrieval approach are becoming more serious, for example: this process islabor-intensive, time-consuming and subjectivity of the description. To overcomethese problems, a content-based image retrieval approach was born. In CBIR, theimage conte ...