Tensor Oriented No-Reference Light Field Image Quality Assessment

Abstract

Light field image (LFI) quality assessment is becoming more and more important, which helps to better guide the acquisition, processing and application of immersive media. However, due to the inherent high dimensional characteristics of LFI, the LFI quality assessment turns into a multi-dimensional problem that requires consideration of the quality degradation in both spatial and angular dimensions. Therefore, we propose a novel Tensor oriented No-reference Light Field image Quality evaluator (Tensor-NLFQ) based on tensor theory. Specifically, since the LFI is regarded as a low-rank 4D tensor, the principle components of four oriented sub-aperture view stacks are obtained via Tucker decomposition. Then, the Principal Component Spatial Characteristic (PCSC) is designed to measure the spatialdimensional quality of LFI considering its global naturalness and local frequency properties. Finally, the Tensor Angular Variation Index (TAVI) is proposed to measure angular consistency quality by analyzing the structural similarity distribution between the first principal component and each view in the view stack. Extensive experimental results on four publicly available LFI quality databases demonstrate that the proposed Tensor-NLFQ model outperforms state-of-the-art 2D, 3D, multi-view, and LFI quality assessment algorithms.Index Terms-Light field, image quality assessment, objective model, tensor theory, angular consistency.

Image Quality Evaluation of Light Field Image Based on Macro-Pixels and Focus Stack

“…LFI FR IQA metrics include the algorithms in Min et al ( 2020 ) and Meng et al ( 2020 ). LFI NR IQA metrics include BELIF (Shi et al, 2019 ), Tensor-NLFQ (Zhou et al, 2020 ), and VBLIF (Xiang et al, 2020 ).…”

“…As an LF image can be regarded as a low-rank 4D tensor, Shi et al ( 2019 ) adopted the tensor structure of the cyclopean image array from the LF to explore the angular-spatial characteristic. Zhou et al ( 2020 ) used tensor decomposition of view stack in four directions to extract the spatial-angular features. To explore the angular-spatial characteristics of LF images, Min et al ( 2020 ) averaged the structural matching degree of all viewpoints to compute the spatial quality and analyzed the amplitude spectrum of near-edge mean square error along viewpoints to express the angular quality.…”

No-Reference Deep Quality Assessment of Compressed Light Field Images

“…Besides, several no-reference LF image quality assessment (NR-LFIQA) methods had also been proposed. Shi et al propose three NR-LFIQA models, including BELIF [13], NR-LFQA [14], and Tensor-NLFQ [15]. BELIF [13] uses binocular vision features of LF images to predict the quality.…”

“…NR-LFQA [14] extracts naturalness of LF cyclopean image to measure spatial quality, calculates the distribution of gradient direction map of Epipolar Plane Image (EPI) to measure angular consistency distortion, and uses weighted local binary to capture local angular consistency degradation feature. Based on BELIF, Tensor-NLFQ [15] exploits angular consistency of each direction and takes the influence of luminance and chrominance on perceptual quality into consideration. VBLFI [16] uses natural scene statistics (NSS) features, and energy features extracted from mean difference image (MDI) and sub-aperture images (SAIs) in the curvelet domain to predict LF image quality.…”

A Measurement for Distortion Induced Saliency Variation in Natural Images

“…We are facing an unprecedented situation where end-users expect high quality visual experiences. Hence, image quality assessment (IQA) has become a popular research topic in both academia and industry [5]- [12]. Recently, a significant approach in IQA is to integrate saliency to its objective algorithms [13]- [18].…”