A novel objective non-reference metric for multimodal image sensor fusion is presented. Using the joint segmentation of the input images and entropy of the regions as the priority, an `ideal' fused image is obtained. The metric is based on computing how accurately the important regions from the input images are transferred into the fused image. Experiments have shown that the obtained metric values correspond better to the subjective quality of the fused images than other state-of-the-art fusion metrics.
Publisher: Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Rose publication type: Journal article