Image Processing Apparatus Depth Accuracy via Surrounding Area Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing technologies face challenges in accurately determining depth values from captured images, particularly in situations where the distance between image capturing devices is small, leading to reduced accuracy and reliability, especially in areas with low contrast.
Innovation Solution
An image processing apparatus and method that derive and improve distance information from an in-focus plane by obtaining captured images, calculating evaluation values for pixel accuracy, and generating new distance information based on surrounding area data, thereby enhancing the accuracy of depth perception.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If the distance between two image capturing apparatuses is reduced, then the device complexity is reduced, but the measurement precision of depth values deteriorates
Solution Approach 1:
The patent combines multiple information sources (parallax information from multiple capturing apparatuses, luminance information from the captured image, and gradient information) to compute depth values. This merging of information sources compensates for the reduced baseline distance, maintaining depth accuracy even when the distance between capturing apparatuses is small.
Solution Approach 2:
The patent introduces an intermediary processing step that uses luminance information and gradient calculations as intermediate data to refine depth values. This intermediary processing layer helps recover depth accuracy that would otherwise be lost due to the reduced distance between capturing apparatuses.
2Productivity
If the stereo method is used to obtain depth values, then the productivity of depth map generation is improved, but the reliability of depth values in low contrast areas deteriorates
Solution Approach 1:
The patent changes the parameters used for depth calculation by incorporating luminance information and gradient magnitudes alongside parallax information. This parameter expansion allows the system to maintain high productivity while improving reliability in low contrast areas by using additional image characteristics.
Solution Approach 2:
The patent performs preliminary calculations of gradient information and luminance characteristics before final depth value determination. This preliminary action prepares additional data that can be used to correct or supplement parallax-based depth values, especially in areas where parallax alone is unreliable.
3Device complexity
If only depth values are used to identify noisy pixels, then the device complexity is reduced, but the measurement precision of depth values deteriorates due to low reliability in flat portions
Solution Approach 1:
The patent changes the basis for noise identification from using only depth values to using a combination of gradient magnitude and luminance information. This parameter change enables more accurate noise detection, particularly in flat portions of images where depth values alone are unreliable.
Solution Approach 2:
The patent uses gradient magnitude and luminance information as intermediary indicators to assess the reliability of depth values before final processing. These intermediary measures provide additional criteria for identifying noisy pixels without significantly increasing device complexity.
Data Source
AI summary
An image processing apparatus, comprises: an obtainment unit that obtains a captured image including a plurality of pixels, and distance information relating to a distance from an in-focus plane and corresponding to each of the plurality of pixels; a deriving unit that derives an evaluation value that indicates an accuracy of the distance information and is based on values of the plurality of pixels; and a generation unit that newly generates, based on the distance information and the evaluation value, distance information for each of the plurality of pixels, wherein the generation unit generates distance information corresponding to an area of interest including at least one pixel, based on distance information of a surrounding area of the area of interest.