HDRI

Ghost Detection in High Dynamic Range Images (HDRI)

A photograph taken with a conventional camera cannot capture the whole dynamic range of real scenes which varies over several orders of magnitude. As a consequence, some regions of the scene will be under- or over-exposed and appear saturated in the image.

To enlarge the dynamic range spanned by conventional cameras a very interesting and powerful technique has been developed in the last few years: high dynamic range imaging. The obtained images are called high dynamic range (HDR) images and represent the scene more faithfully than conventional low dynamic range (LDR) images.

A classic approach for obtaining an HDRI with a conventional camera is to take a sequence of images of the same scene with different exposure times, and combine them to a single radiance map. This multiple exposures technique suffers from two main problems:

Misalignment: if the camera moves during the time of capture, the images will be misaligned and the combined HDRI will look blurry.

Ghosting: if there are moving objects while capturing the sequence of images, these objects will appear in different locations in the combined HDRI, creating what is called ghost or ghosting artifacts.

The first problem can be solved easily by placing the camera on a tripod or by using an image registration method to align the differently exposed images. On the contrary, the second problem is a severe limitation of the multiple exposures technique since motion is hardly avoidable in outdoor environments.

Evaluation of ghost detection methods

Detecting and removing ghosting artefacts is an important issue for automatic generation of HDR images in dynamic environments.

Different algorithms have been developed to solve the ghost problem.

An up-to-date review of the recently proposed methods for ghost-free HDR image generation is given in the recent review paper in [2].

A classification and comparison of the reviewed methods is reported to serve as a useful guide for future research on this topic.

Dataset

In order to facilitate the evaluation of ghost detection algorithms, a sequence of images showing ghost artefacts with different characteristics is designed and made available.

It contains a manually segmented ground-truth ghost map for quantitative analysis.

The sequence of images used in the review paper can be obtained here.

Please cite the following paper if you use this dataset in your research:

A. Srikantha, D. Sidibé, "Ghost Detection and Removal for High Dynamic Range Images: Recent Advances", Signal Processing: Image Communication, 27(6), pp. 650-662, 2012

link to the paper: link

A simple and effective method

In the paper [3], an elegant ghost detection approach based on SVD is proposed.

The method, though simple, shows good ghost detection results.

Dataset

The sequence of images used in the review paper can be obtained here.

Please cite the following paper if you use this dataset in your research:

A. Srikantha, D. Sidibé, "Ghost Detection and Removal for High Dynamic Range Images: Recent Advances", Signal Processing: Image Communication, 27(6), pp. 650-662, 2012

References

[1] D. Sidibé, W. Puech and O. Strauss "Ghost Detection and Removal in High Dynamic Range Images", EUSIPCO 2009. 17th European Signal Processing Conference, Glasgow, Scotland, August 2009. (link)

[2] A. Srikantha, D. Sidibé, "Ghost Detection and Removal for High Dynamic Range Images: Recent Advances", in Signal Processing: Image Communication, 2012. (link)

[3] A. Srikantha, D. Sidibé, F. Meriaudeau, "An SVD-Based Approach for Ghost Detection and Removal in High Dynamic Range Images", To appear in 21st International Conference on Pattern Recognition, ICPR 2012. (link)