3D Depth Continuity Grouping for Stixel Data Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods, such as Stixel representation, struggle to effectively describe depth continuity in three-dimensional structures, leading to inefficiencies in data representation and difficulty in distinguishing between constant and varying depth changes, particularly at intersections or corners.
Innovation Solution
An image processing device and method that calculates three-dimensional coordinates from range images, detects reference surfaces, identifies object regions, calculates depth changes, and groups vertical lines with consistent depth changes to reduce data while maintaining depth continuity representation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If point cloud data is divided into strip-shaped segments (Stixel) to reduce data amount, then data compression is improved, but depth continuity representation deteriorates
Solution Approach 1:
The patent divides the point cloud data into vertical line segments (Stixel) along the vertical direction, grouping points that share the same horizontal coordinates. This segmentation reduces data complexity while maintaining the essential depth information through the depth continuity flag that indicates whether depth values change along vertical lines.
Solution Approach 2:
The patent introduces a depth continuity flag parameter that changes state based on whether depth values remain constant along vertical lines. This parameter transformation allows the system to efficiently represent depth continuity information without significantly increasing data volume, resolving the contradiction between compression and information retention.
2Productivity
If Stixel representation is used to structure point cloud data, then data processing efficiency is improved, but distinction between constant and varying depth changes deteriorates
Solution Approach 1:
The patent applies different representation qualities to different regions of the point cloud data. For vertical lines with constant depth, a simplified representation is used, while for lines with varying depth, more detailed information is retained. This local differentiation improves processing efficiency while maintaining precision where needed.
Solution Approach 2:
The depth continuity flag serves as a parameter that changes based on the local characteristics of depth variation. When depth is constant along a vertical line, the flag indicates this state, enabling efficient processing. When depth varies, the flag changes to preserve detailed information, thus maintaining measurement precision.
3Quantity of substance
If vertical lines are grouped by depth similarity, then data compression is improved, but ability to detect depth discontinuities (intersections/corners) deteriorates
Solution Approach 1:
The depth continuity flag acts as an intermediary indicator that bridges compressed data representation and reliable discontinuity detection. This flag provides additional information about depth characteristics without significantly increasing data volume, enabling reliable detection of intersections and corners even in compressed representations.
Solution Approach 2:
The patent retains more depth information than strictly necessary for basic compression by including the depth continuity flag. This partial retention of excessive information ensures that depth discontinuities can be reliably detected while still achieving meaningful compression ratios.
Data Source
AI summary
Image processing device (100) includes range image obtaining unit (110), reference surface detector (130), object region detector (140), depth calculator (150), and grouping unit (160). Range image obtaining unit (110) obtains a range image. Reference surface detector (130) detects a reference surface having a predetermined height in the range image. Object region detector (140) detects an object region existing on the reference surface. Depth calculator (150) decides a depth in each vertical line of the object region. Grouping unit (160) calculates a change amount of depth between adjacent vertical lines in the object region, groups a set of vertical lines in which the change amount of depth is less than or equal to a predetermined threshold, and describes each group in a predetermined data format.