This thesis presents methods to improve instance search and enhance user performance while browsing unstructured video collections. Through the use of computer vision and information retrieval techniques, we propose novel solutions to analyse visual content and build a search algorithm to address the challenges of visual instance search, while considering the constraints for practical applications. Firstly, we investigate methods to improve the effectiveness of instance search systems for finding object instances which occurred in unstructured video content. Using the bag of feature framework, we propose a novel algorithm to use the geometric correlation information between local features to improve the accuracy of local feature matching, thus improve the performance of instance search systems without introducing much computation cost. Secondly, we consider the scenario that the performance of instance search systems may drop due to the volume of visual content in large video collections. We introduce a search algorithm based on embedded coding to increase the effectiveness and efficiency of instance search systems. And we participate in the international video evaluation campaign, TREC Video Retrieval Evaluation, to comparatively evaluate the performance of our proposed methods. Finally, the exploration and navigation of visual content when browsing large unstructured video collections is considered. We propose methods to address such challenges and build an interactive video browsing tool to improve user performance while seeking interesting content over video collections. We construct a structured content representation with similarity graph using our proposed instance search technologies. Considering the constraints related to real world usability, we present a flexible interface based on faceted navigation to enhance user performance when completing video browsing tasks. This thesis shows that user performance can be enhanced by improving the effectiveness of instance search approaches, when seeking information in unstructured video collection. While covering many different aspects of improving instance search in this work, we outline three potential directions for future work: advanced feature representation, data driven rank and cloud-based search algorithms.