Computational Vision and Imaging Technology Research Group

Research projects

Methods and metrics: Evaluation of Image Quality in Still Image Compression

Elizabeth’s research explores the various approaches used in the measurement and evaluation of the image quality, in this case of JPEG2000 lossy compressed images.

A focus of the investigation is a performance evaluation of different types of image quality metrics used to model and predict quality as perceived quality by human observers, as well as the success of methods recommended for subjective image quality evaluation. Of particular interest are scene dependency and its relationship with image quality: the effect of scene content on the performance of the algorithm and on the perceived quality of the images

The work attempts to address the following research questions:

  1. How well does the JPEG 2000 compression algorithm perform in terms of image quality and image fidelity in a general imaging context, especially in relation to other equivalent lossy compression algorithms?
  2. What are the main approaches to objective image quality assessment for compressed images and which of them are able to produce results most consistent with subjective quality evaluation?
  3. Which type of metric is best able to predict scene dependency in the results of an image quality assessment and why?

The questions are been tackled using a series of psychophysical investigations dealing with the definition of the acceptability and the perceptibility of artefacts introduced by image compression as well as studies on existing and novel metrics for objective quantification of the compressed image quality.

Doctoral Researchers

People

The Academic staff and Doctoral Researchers who make up the group.

Contact

Get in touch to get more information about Computational Vision Imaging Technology.