Authors : N. Sivakumar and K. Helenprabha
Abstract: Multiple acquisition schemes are utilized to acquire complementary informative description associated with an object. Image fusion improve the features of an image (depicting comprehensive description of the object) acquired from different acquisition schemes. The existing homogeneous schemes fail to assign membership to voxels thereby yielding noisy boundaries in the core. In this work, a novel scheme to optimize the kernel is done through class separability criterion using OpenCV tools on hardware (s3c2440a processor) and the algorithms are scalable and generic. PCA based fusion intrinsic dimensionality reduction that uses optimal set of parameters required to account the experiential properties of the data is used. The Bigram Proximity Matrix captures the number of times a feature pair occurs with its columns and rows representing the features of CT and MRI scanned images. The metrics used to evaluate the performance includes MSE, PSNR and Cross entropy. Improved results are achieved with minimum MSE, minimum cross entropy and maximum PSNR.
N. Sivakumar and K. Helenprabha, 2016. Fusion of CT and MRI Images Using OpenCV with FL2440 Hardware. Asian Journal of Information Technology, 15: 1936-1944.