Estimating Depth-Salient Edges and Its Application to Stereoscopic Image Quality Assessment.

Area of Science:

• Computer Vision
• Image Processing
• Human-Computer Interaction

Background:

• Human visual attention prioritizes salient image regions.
• Stereoscopic 3D (S3D) images incorporate depth information, influencing visual perception.
• Depth cues, specifically image edges, play a role in S3D depth perception.

Purpose of the Study:

• To propose a systematic approach for depth saliency estimation in S3D images.
• To introduce Salient Edges with respect to Depth perception (SED) for localizing depth-salient edges.
• To demonstrate the utility of SED in full-reference stereoscopic image quality assessment (S3D IQA).

Main Methods:

• Developed the Salient Edges with respect to Depth perception (SED) method.
• Utilized gradient magnitude and inter-gradient maps for structural similarity prediction.
• Estimated luminance quality from 2D saliency and gradient maps, then refined with SED for depth quality.

Main Results:

• SED effectively localizes depth-salient edges in S3D images.
• The proposed S3D IQA metric combines luminance and depth quality.
• The metric achieved competitive performance across seven S3D IQA databases, with state-of-the-art results on three.

Conclusions:

• Depth saliency estimation using SED is a significant factor in S3D image quality.
• The proposed method provides a robust approach for evaluating S3D image quality.
• SED contributes to advancing the field of stereoscopic image quality assessment.