Image Composing Device Memory Bandwidth Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image composing devices for vehicles face issues with dynamic viewpoint changes, leading to image loss and unnatural displays due to insufficient memory access bandwidth, causing previous line images to be displayed instead of the correct composed image.
Innovation Solution
The image composing device replaces camera images that cannot be read out due to memory access shortages with alternative data, such as bit map data, ensuring that the composed image is always accurately displayed by controlling memory access within the available bandwidth.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If dynamic viewpoint changing is implemented in image composition, then the adaptability to different driving situations is improved, but the reliability of image display deteriorates due to image loss and unnatural displays
Solution Approach 1:
The patent pre-calculates and stores conversion tables for multiple viewpoint positions before runtime. When viewpoint changing is needed, the system simply retrieves the pre-prepared conversion table corresponding to the target viewpoint, avoiding real-time calculation that would consume excessive memory bandwidth and cause image loss.
Solution Approach 2:
The patent implements dynamic viewpoint changing by selecting from multiple pre-calculated conversion tables based on the current driving situation. The system dynamically switches between different viewpoint configurations (e.g., front view, rear view, side views) according to vehicle speed, steering angle, and other driving parameters, maintaining adaptability while ensuring reliable display.
2Manufacturing precision
If interpolation is used to improve image quality, then the manufacturing precision of the composed image is improved, but the quantity of memory access increases beyond the available bandwidth
Solution Approach 1:
The patent performs image interpolation and composition operations in advance, storing the results in pre-calculated conversion tables. During actual operation, the system only needs to retrieve pre-processed image data and compose the final view, dramatically reducing real-time memory access requirements while maintaining high image quality through interpolation.
Solution Approach 2:
The patent divides the image processing into separate stages: pre-processing (interpolation and initial composition) and real-time composition. By segmenting the computationally intensive interpolation operation from the time-critical display operation, the system can achieve high image quality without exceeding memory bandwidth constraints during dynamic viewpoint changing.
3Device complexity
If the accessed band in frame memory is limited, then the device complexity is reduced, but the reliability of image read-out deteriorates due to insufficient bandwidth
Solution Approach 1:
The patent pre-calculates and stores conversion tables that map input image coordinates to output image coordinates for various viewpoint positions. This allows the system to use simple memory access patterns (reading pre-organized data) rather than complex real-time calculations, ensuring reliable image read-out within limited bandwidth constraints.
Solution Approach 2:
The patent creates pre-processed copies of image data in the form of conversion tables during off-line or initial processing. These copied and pre-organized data structures enable fast, reliable retrieval during runtime without requiring high memory bandwidth, as the system reads ready-to-use converted data rather than performing conversions in real-time.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An object of this invention is to provide an image composing device and image composing method which can exactly display a composed image even if a virtual viewpoint is dynamically changed. A coordinate conversion table indicative of the correlation between an input image and a distortion-corrected image and the parameter on a virtual viewpoint position necessary to create a composed image are prepared; on the basis of the coordinate conversion table and the parameter on the virtual viewpoint point, virtual viewpoint conversion data indicative of the correlation between the input image and the virtual viewpoint are created; write of camera images for a frame memory (15) is executed and also read of the camera images based on the virtual viewpoint conversion data is executed using the remaining accessed band in the frame memory (15), thereby to create the composed image; and if no camera image can be read out owing to shortage in the accessed band in said frame memory (15), the camera image which could not read out from the frame memory (15) is replaced by any bit map data.