Blind Image Quality Assessment Using Local Variant Patterns

This paper introduces a new blind image quality assessment (BIQA) metric using texture analysis. The method adopts two texture operators to select image texture information. The first operator is the Local Binary Pattern (LBP), an effective texture operator that is extensively adopted for texture analysis. The second operator is proposed as an extension of LBP. The proposed operator, the Local Variant Pattern (LVP), extracts local energy information. Energy information is particularly important for BIQA metrics because image distortions modify the energy of the textures. Histograms of the LBP and LVP outputs are used as features in a random forest regression algorithm. The proposed method surpass other state-of-the-art BIQA method, as results demonstrate.