Performance comparison of decision fusion strategies in BMMF-based image quality assessment

The block-based multi-metric fusion (BMMF) is one of the state-of-the-art perceptual image quality assessment (IQA) schemes. With this scheme, image quality is analyzed in a block-by-block fashion according to the block content type (i.e. smooth, edge and texture blocks) and the distortion type. Then, a suitable IQA metric is adopted to evaluate the quality of each block. Various fusion strategies to combine the QA scores of all blocks are discussed in this work. Specifically, factors such as quality scores distribution and the spatial distribution of each block are examined using statistics methods. Finally, we compare the performance of various fusion strategies based on the popular TID database.