Supheakmungkol Sarin

On the design and exploitation of user's personal and public information for semantic personal digital photograph annotation

Automating the process of semantic annotation of digital personal photographs is a crucial step towards efficient and effective management of this increasingly high volume of content. However, this is still a highly challenging task for the research community. This paper proposes a novel solution. Our solution integrates all contextual information available to and from the users, such as their daily emails, schedules, chat archives, web browsing histories, documents, online news, Wikipedia data, and so forth. We then analyze this information and extract important semantic terms, using them as semantic keyword suggestions for their photos. Those keywords are in the form of named entities, such as names of people, organizations, locations, and date/time as well as high frequency terms. Experiments conducted with 10 subjects and a total of 313 photos proved that our proposed approach can significantly help users with the annotation process. We achieved a 33% gain in annotation time as compared to manual annotation. We also obtained very positive results in the accuracy rate of our suggested keywords.

Leveraging Features from Background and Salient Regions for Automatic Image Annotation

In this era of information explosion, automating the annotation process of digital images is a crucial step towards efficient and effective management of this increasingly high volume of content. However, this still is a highly challenging task for the research community. One of the main bottlenecks is the lack of integrity and diversity of features. We propose to solve this problem by utilizing 43 image features that cover the holistic content of the image from global to subject, background and scene. In our approach, salient regions and the background are separated without prior knowledge. Each of them together with the whole image are treated independently for feature extraction. Extensive experiments were designed to show the efficiency and the effectiveness of our approach. We chose two publicly available datasets manually annotated with diverse nature of images for our experiments, namely, the Corel5K and ESP Game datasets. We confirm the superior performance of our approach over the use of a single whole image using sign test with p-value < 0.05. Furthermore, our combined feature set gives satisfactory performance compared to recently proposed approaches especially in terms of generalization even with just a simple combination. We also obtain a better performance with the same feature set versus the grid-based approach. More importantly, when using our features with the state-of-the-art technique, our results show higher performance in a variety of standard metrics.

Classification and quality assessment of high quality digital photographs

We are interested in high quality photographs. This paper outlines our research proposal for the tasks of classification and quality assessment. We address these challenges by exploring the aesthetics from the combined perspectives of the artists and the photographers. We propose to use the aesthetic primitives of images for visualization as a guideline for high and low-level image feature extraction and to classify this high quality content into six creative exposure themes, which are commonly followed by the professional photographers. Then, we suggest to evaluate the quality of the photograph accordingly to these themes. We solve the problems using statistical modeling and learning approach.