Objective image quality assessment of 3D synthesized views
作者：
Highlights：
• This paper presents a new full- reference objective quality metric dedicated to artifacts detection in 3D synthesized views.
• The proposed metric is based on the comparison of statistical features of wavelet subbands of the original image and the DIBR-based synthesized image.
• “Shifting-resilience” is granted by the use of a registration algorithm.
• The final quality score is weighted depending on the presence of “skin-pixels” based on the assumption that a human observer is more sensitive to impairments affecting human subjects.
• Experimental tests show that the proposed method outperforms the 2D conventional and DIBR-synthesized views dedicated quality metrics under test.
摘要
Highlights•This paper presents a new full- reference objective quality metric dedicated to artifacts detection in 3D synthesized views.•The proposed metric is based on the comparison of statistical features of wavelet subbands of the original image and the DIBR-based synthesized image.•“Shifting-resilience” is granted by the use of a registration algorithm.•The final quality score is weighted depending on the presence of “skin-pixels” based on the assumption that a human observer is more sensitive to impairments affecting human subjects.•Experimental tests show that the proposed method outperforms the 2D conventional and DIBR-synthesized views dedicated quality metrics under test.
论文关键词：Objective image quality,3D synthesized views,DIBR,Skin detection
论文评审过程：Received 27 May 2013, Revised 20 October 2014, Accepted 20 October 2014, Available online 4 November 2014.
论文官网地址：https://doi.org/10.1016/j.image.2014.10.005
