Three-layer Medical Image Fusion with Tensor-based Features

Abstract Replacing Computed tomography (CT) with magnetic resonance image (MRI), MRI- positron electron tomography (PET) or MRI- single photon emission computed tomography (SPECT) imaging might have further advantages due to higher soft-tissue contrast of brain structure and lower dose absorbed by the patient. In this paper, a new three-layer (intensity, detail, and base layers) medical image fusion method with differential features for gray and pseudo-color images is proposed. The proposed method includes three steps. At the first step, differential feature by structure tensor is used to decompose the anatomical MRI medical image into its three-layer image representation. On the other side, differential feature by color tensor is adopted to decompose the functional PET or SPECT medical image into its three-layer image representation. At the second step, spatial frequency metric is proposed to combine the decomposed intensity layers and detail layers and absolute maximum is defined as the image fusion rule of the base layers. At the third step, the fused image is obtained by the addition of the fused intensity layer, the fused detail layer, and the fused base layer. The superiority of the proposed method is demonstrated by subjective and objective evaluation on experimental results.