QUALITAS: Image Quality Assessment for Stereoscopic Images

Perceptual stereoscopic image quality assessment method with tensor decomposition and manifold learning

“…To effectively describe the binocular fusion and rivalry properties, we defined a pooling strategy based on local energy ratio similar to [10] to determine the weights for left and right views.…”

“…To verify the effectiveness of the proposed method, we comprehensively compared the proposed method, TDML, with 13 state-of-the-art IQA metrics. Particularly, eight full-reference SIQA metrics (Benoit's [4], You's [5], Chen's [6], Shao's [29], Jiang's [7], QUALITAS [10], Geng's [11] and Fan's [17]) were used for reference; and five classical IQA metrics peak signal to noise ratio (PSNR), structural similarity (SSIM), multi-scaled structural similarity (MS-SSIM) [30], visual information fidelity (VIF) [31], and universal quality index (UQI) [32]) were extended to the 3D case. Note that, for the above five classical IQA metrics, referring to [9], left and right quality scores were calculated separately and then integrated by the binocular integration behaviours to obtain the final estimate.…”

“…Lin et al [9] integrated binocular fusion and binocular frequency for designing SIQA metric, denoted as frequency-integrated metrics. Fernandez et al [10] presented a novel method: QUALITAS, which was inspired by some features of the HVS, such as contrast sensitivity, response to visual disparity and perception of distance. QUALITAS employed a penalisation depending on the corresponding wavelet energy so that dynamic parameters were considered for two views, yielding overall estimates.…”