Vision for Fast Image Matching and Search

Back to Verticals page

In this vertical we aspire to develop efficient algorithms and frameworks that enable rapid identification and retrieval of visual content. By leveraging cutting-edge techniques, our goal is to streamline image-matching processes, fostering quicker and more accurate search capabilities across diverse applications.

a. Correlation for Fast Matching and Search

This study focuses on the optimization of autocorrelation for rapid matching and search. With an emphasis on refining auto-correlation techniques, our objective is to enhance their efficiency in various applications such as image recognition, signal processing, and data retrieval. By exploring strategies to expedite matching and search operations, this research aims to contribute to the development of computational methods that play a pivotal role in diverse domains.

Related Publications

● Optimizing Autocorrelation for Fast Target Search in Large Search Space, arXiv, 2014

● On Optimizing Autocorrelation for Fast Template Matching Through Transitive Elimination, arXiv, 2014
● Exploiting transitivity of correlation for fast template matching, IEEE Transactions on image processing (TIP), 2010

b Early Termination to Speed-up Matching

In this study, our focus is on early termination-based methods aimed at accelerating detection and matching. We explore algorithms that enable efficient termination of these operations, enhancing speed and optimizing resource utilization. Through the investigation of early termination techniques, our goal is to contribute advancements in detection and matching methodologies, promoting faster and more resource-efficient algorithms.

Related Publications

● Correlation-coefficient based Fast Template Matching Through Partial Elimination, IEEE Transactions on image processing (TIP), 2012
● Early terminating algorithms for Adaboost-based detectors, IEEE International Conference on Image Processing (ICIP), 2009
● Early termination algorithms for correlation coefficient-based block matching, IEEE International Conference on Image Processing (ICIP), 2007

c. Video coding with linear compensation (VCLC)

In this study, we explore methods involving the use of the difference between the current block and its first-order linear estimate from the best-matching block. Motivated by frequent brightness and contrast changes observed in real videos, this approach aims to address dynamic content more effectively. These future- oriented techniques are anticipated to contribute to improved video compression methodologies.

Related Publications

● Video Coding With Linear Compensation (VCLC), International Conference on Communications (ICC), 2007

d. Fast Video to Reference Image Alignment

In this study, we explore the utilization of the inherently strong spatial correlation found in digital images of natural scenes. Traditionally leveraged in still and moving picture coding, we aim to extend this principle to enhance the efficiency of video-to-reference image alignment algorithms.

Related Publications

● Exploiting local auto-correlation function for fast video to reference image alignment, IEEE International Conference on Image Processing (ICIP), 2008
● Exploiting inter-frame correlation for fast video to reference image alignment, Asian Conference on Computer Vision (ACCV), 2007

Vision for Fast Image Matching and Search

AI for Multimodal Learning

Back to Verticals page