Stereoscopic Image Depth Correction for Edge Consistency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional stereoscopic image generation techniques face challenges in creating consistent parallax images, particularly at the edges of the screen, leading to broken or inconsistent views when shifting pixels based on depth values, as there are no reference pixels outside the screen to accurately estimate and apply parallax.
Innovation Solution
A stereoscopic image converting apparatus that includes a depth estimating module to generate a depth map, a right and left end depth corrector to adjust depth values near the screen edges to match the screen surface, and a parallax image generator to shift pixels based on corrected depth values, preventing edge inconsistencies by ensuring depth values align with the screen surface.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If depth values at the right and left ends of the input image are made distant from the screen surface to generate parallax images, then the parallax effect is enhanced, but the ends of the screen appear broken and the parallax images become inconsistent
Solution Approach 1:
The patent applies different depth value treatments to different regions of the image. Specifically, pixels within a predetermined distance from the right and left ends have their depth values corrected to be close to the screen surface, while other pixels maintain their original depth values. This local differentiation resolves the contradiction by ensuring edge consistency without compromising the overall parallax effect.
Solution Approach 2:
The patent performs preliminary correction of depth values for edge pixels before generating the final parallax images. By pre-adjusting the depth values of pixels near the edges to be close to the screen surface, the system prevents edge breakage from occurring in the first place, ensuring consistent parallax images without requiring post-processing corrections.
2Reliability
If pixels are shifted based on depth values to generate multi-parallax images, then the stereoscopic effect is improved, but reference pixels outside the screen are unavailable leading to edge breakage
Solution Approach 1:
The patent introduces a spatially varying processing approach where pixels within a predetermined distance from the image ends undergo depth value correction, while other pixels are processed normally. This local quality differentiation ensures that edge pixels have depth values close to the screen surface, preventing edge breakage during the pixel shifting process without affecting the overall stereoscopic effect.
3Manufacturing precision
If depth values are corrected for pixels near the screen edges, then edge consistency is improved, but the processing complexity increases
Solution Approach 1:
The patent segments the image processing into distinct regions: pixels within a predetermined distance from the right and left ends receive depth value correction, while other pixels maintain their original depth values. This segmentation approach allows the system to apply complex processing only where necessary (at the edges) while keeping the majority of the image processing simple, thus balancing edge precision with overall processing efficiency.
Data Source
AI summary
According to one embodiment, a stereoscopic image converting method includes: generating a first depth map comprising depth values of pixels of an input video, wherein the depth values are estimated based on an amount of a characteristic of the input video, and wherein the depth values are represented by pixel shading; generating a second depth map by correcting the depth values of pixels within a distance from each of right and left ends of the depth map to become closer to a depth value corresponding to a screen of a display not protruding forward or recede rearward in stereoscopic viewing as the pixels within the first distance are positioned closer to the each of the right and left ends of the depth map; and generating a multi-parallax image by shifting pixels of the input video by an amount horizontally based on the depth values of the second depth map.