Adaptation of mutual information measure by using image gradient information

Tan Chye Cheah, S. Anandan Shanmugam, Li-Minn Ang

    Research output: Contribution to journalArticlepeer-review

    2 Citations (Scopus)

    Abstract

    In recent years, mutual information has developed as a popular similarity measure especially in multimodality image registration. However, based on Shannon entropy, it only takes statistical information between corresponding individual pixels into consideration and ignores the spatial information contained in the images such as edges and corners that might be useful in the image registration. As including spatial information into mutual information can help in handling misregistration where the conventional mutual information is challenged. Thus we propose the adaptation of mutual information measure which incorporates the gradient information. The new mutual information value is calculated from the new image description based on the combination of image gradient value and original intensity value of the images. Salient pixels in the regions with high gradient value contribute more in the estimation of mutual information of image pairs being registered. We then compare the registration result with the existing normalized mutual information and mutual information using image gradient alone. The experimental results show that the new method yield better registration accuracy and it is more robust to noise than normalized mutual information.
    Original languageEnglish
    Pages (from-to)313-319
    Number of pages7
    JournalJournal of Medical Imaging and Health Informatics
    Volume2
    Issue number3
    DOIs
    Publication statusPublished - Sept 2012

    Fingerprint

    Dive into the research topics of 'Adaptation of mutual information measure by using image gradient information'. Together they form a unique fingerprint.

    Cite this