This research on video anomaly detection method based on wake motion descriptors. We are developing a method that analyses the motion characteristics of the video data, on a video volume-by-video volume basis, by computing the wake left behind by moving objects in the scene. It then probabilistically identifies those never previously seen motion patterns in order to detect anomalies. The method also considers the perspective of the scene to compensate for the relative change in an object's size introduced by the camera's view angle.  To this end, a perspective grid is proposed to define the size of video volumes for anomaly detection.

The HEVC reference software includes a uniform reconstruction quantization (URQ) method. This block level quantization technique does not take into account the importance of transform coefficients in terms of signal reconstruction. This project focuses on the design of transform coefficient level quantization techniques based on soft thresholding in which the quantization parameter (QP) of a transform coefficient is altered according to its importance in the signal reconstruction process as well as according to the overall energy of the corresponding transform block (TB).

PhD student: Olly Styles

Project: Graph-based signal processing for pathology image coding

Nagaoka University of Technology, Japan

March-May 2017

Instituto Politecnico Nacional, Mexico

March-August 2016

This research focuses on constructing an automatic image analysis system for improving the accuracy of diagnosis of bone marrow diseases. Hematopathologists analyze and diagnose the stages of bone marrow diseases manually based on their clinical experiences. However, this manual approach can be tedious and prone to errors. We are developing computational techniques to assist hematopathologists in identifying features that are indicative of disease.

PhD student: Guannan Li

Uterine Natural Killer (uNK) cells are immune cells found in the human female uterus lining. Normally, these cells make up no more than 5% of all cells in the womb lining. Recently, it has been shown that there are abnormally high numbers of uNK cells in the uterus of women who suffer from recurrent miscarriages. In this project, we developed computer-assisted diagnosis methods to detect over-presence of uNK cells in histopathology images of endometrial biopsy stained with Haematoxylin and CD56 (which stains uNK cells brown when used with DAB staining).

Funding: King Adbulaziz University (KAU) and The Saudi Arabian Cultural Bureau (SACB), London

Parametric active contours, or snakes, are curves defined within an image domain that can move under the influence of internal forces from within the curve itself and external forces computed from the image data. Weak edges are one of the main challenges for snakes, as they are usually incapable of detecting them. In this project, we are developing methods to overcome this challenge while minimizing the initialization process associated with these curves.

As the only biometric authentication method from a distance, gait recognition has been studied for years. However, the performance of existing algorithms is still not good enough for real world cases. In this project, we are analyzing the impact of low resolution and noise in gait silhouette and the corresponding gait energy images (GEI) on the identification process. We are designing experiments intended to simulate more realistic scenarios, for example resolution differences between gallery and probe GEIs. Our experiments are aimed at simulating segmentation or image restoration differences on gait silhouettes. Recognition under these differences varies significantly. In order to tackle this problem, we are investigating the applicability of various dimensionality reduction methods that can accurately separate classes after projection, while keeping data structure.

Phd student: Alasdair Thomason

Funding: EPSRC

This research is focused on predictive analytics for next-generation location-aware systems. That is, analysing data collected from location-aware devices, such as smartphones and data loggers, to make predictions about the behavior of individuals and groups.

This project focused on improving safety applications on vehicular ad hoc networks (VANETS). Specifically, it is concerned with reducing the delivery delay of safety messages transmitted within the IEEE 802.11p/WAVE (Wireless Access in Vehicular Environments) Standard.