Blind quality assessment of JPEG compressed images
Pranav Sodhani, Haradhan Chel · 2017 Progress in Electromagnetics Research Symposium - Fall (PIERS - FALL) · 2017
This letter presents a no-reference image quality assessment algorithm for measuring blockiness in JPEG compressed images. The proposed metric dubbed BRIBE (Blind/Referenceless Image Blockiness Estimate) operates in the spatial domain and straightaway targets artifacts which cause visual annoyance instead of explicitly incorporating characteristics of the Human Visual System (HVS). It is observed that the presence of horizontal and vertical image gradients at block boundaries along with block flatness is responsible for quality degradation in JPEG compressed images. Both these factors are mathematically captured and finally weighted to obtain the proposed BRIBE index. Experimental results on four JPEG image quality databases show that BRIBE statistically outperforms existing indices on LIVE IQA database and performs at par with them on other databases. A software release of the model is available at https://sites.google.com/site/blindiqa/bribe.