Psycho-physical limits of interocular blur suppression and its application to asymmetric stereoscopic video delivery

Abstract

Abstract-It is well known that when the two eyes are provided with two views of different resolutions the overall perception is dominated by the high resolution view. This property, known as binocular suppression, is effectively used to reduce the bit rate required for stereoscopic video delivery, where one view of the stereo pair is encoded at a much lower quality than the other. There have been significant amount of effort in the recent past to measure the just noticeable level of asymmetry between the two views, where asymmetry is achieved by encoding views at two quantization levels. However, encoding artifacts introduce both blurring and blocking artifacts in to the stereo views, which are perceived differently by the human visual system. Therefore, in this paper, we design a set of psycho-physical experiments to measure the just noticeable level of asymmetric blur at various spatial frequencies, luminance contrasts and orientations. The subjective results suggest that humans could tolerate a significant amount of asymmetry introduced by blur, and the level of tolerance is independent of the spatial frequency or luminance contrast. Furthermore, the results of this paper illustrate that when asymmetry is introduced by unequal quantization, the just noticeable level of asymmetry is driven by the blocking artifacts. In general, stereoscopic asymmetry introduced by way of asymmetric blurring is preferred over asymmetric compression. It is expected that the subjective results of this paper will have important use cases in objective measurement of stereoscopic video quality and asymmetric compression and processing of stereoscopic video.

Prediction architecture based on block matching statistics for mixed spatial-resolution multi-view video coding

“…According to the suppression theory of binocular vision, the total perceived quality for mixed spatialresolution stereoscopic video is close to the view with the highest quality (the view with full spatial-resolution frames) [2,6]. This is due to the high frequency components (which exist in the full spatial-resolution frames) which compensate the corresponding components in the lower spatial-resolution frames [7]. Asymmetric temporalresolution and asymmetric quality are other alternatives for asymmetric coding.…”

“…Asymmetric temporalresolution and asymmetric quality are other alternatives for asymmetric coding. The former causes flickering artifacts especially when coding sequences, which contain fast object motion, while the latter produces inevitable blocking artifacts when coding videos at low bitrates [7,8]. Still, the mixed spatial-resolution approach provides better perceived quality than other coding approaches when coding multi-view videos at low bitrates [2,9].…”

Visual Quality Assessment of Stereoscopic Image and Video: Challenges, Advances, and Future Trends

“…Therefore, compression artifacts are perceived as local distortions, mainly affecting the 2D image quality of stereoscopic/3D videos. In [27], Silva et al conducted a subjective experiment to compare the impact of asymmetrically distorted stereoscopic videos with blurring and blocking artifacts. The results suggest that perceptual quality is dominated conversely by the higher quality view when the lower quality view is degraded by blur.…”

Asymmetric coding using Binocular Just Noticeable Difference and depth information for stereoscopic 3D

“…A promising technique for coding stereo pairs known as asymmetric coding has focused considerable research efforts [4][5][6][7][8][9][10][11][12][13][14][15][16][17]. This approach is based on the so-called "binocular suppression theory" [3,4], which specifies that if one of the views of a given stereo-pair is altered, the 3D perceived quality will be close to the view with the highest quality [9], provided that the gap between both views does not exceed a threshold [5,15,16].…”

“…This approach is based on the so-called "binocular suppression theory" [3,4], which specifies that if one of the views of a given stereo-pair is altered, the 3D perceived quality will be close to the view with the highest quality [9], provided that the gap between both views does not exceed a threshold [5,15,16]. Depending on how the quality reduction of one of the views is achieved, the asymmetric stereoscopic coding methods can be classified into three categories: (i) spatial resolution reduction (spatial filtering) [7][8][9][10][11][12][13][14], (ii) asymmetric quantization (unequal QP) [5,6,15] and (iii) temporal resolution reduction [17].…”