Multiresolution Image Frame Alignment for Faster Motion Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face significant computational challenges in processing high-resolution image frames, leading to increased processing time and resource consumption for motion estimation and alignment, which affects the accuracy and speed of object detection.
Innovation Solution
Downsize or down-sample image frames to lower resolutions, determining alignment information between smaller sized frames to align higher resolution frames, reducing computation complexity and improving processing speed and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high-resolution image frames are processed directly for motion estimation, then measurement precision is improved, but computing resources and processing time increase significantly
Solution Approach 1:
The patent segments the image processing task by dividing it into multiple resolution levels. First, alignment is performed on downsampled low-resolution versions of image frames to obtain initial alignment parameters. Then, these parameters are used to guide alignment of progressively higher-resolution versions. This segmentation allows the system to achieve accurate alignment without processing all pixels at full resolution simultaneously, thus balancing precision and processing speed.
Solution Approach 2:
The patent introduces a resolution dimension to the image processing problem. Instead of working directly with full-resolution images in a single step, the system processes images at multiple resolution levels (dimensions). By adding this resolution dimension and processing from coarse to fine, the system reduces the computational complexity while maintaining alignment accuracy, effectively resolving the contradiction between precision and speed.
2Measurement precision
If high-resolution image frames are processed directly for motion estimation, then alignment accuracy is improved, but device complexity and computation load increase
Solution Approach 1:
The patent segments the computation into hierarchical levels based on resolution. Each level processes a downsampled version of the image frames, with computation complexity increasing gradually from coarse to fine levels. This segmentation reduces the overall computational burden compared to processing full-resolution frames directly, while still achieving accurate alignment through the cumulative effect of multi-level processing.
Solution Approach 2:
The patent applies partial action by performing alignment operations on downsampled versions of image frames rather than full-resolution frames. The downsampling reduces the number of pixels that need to be processed, thereby reducing computation complexity. The alignment parameters obtained from partial processing of downsampled frames are then used to guide the alignment of higher-resolution frames, achieving good results with reduced computational effort.
3Productivity
If downsampling is applied to reduce computation load, then processing speed is improved, but image quality and detection accuracy may deteriorate
Solution Approach 1:
The patent segments the resolution processing into hierarchical levels. Instead of using a single downsampled resolution, it processes multiple levels from coarse to fine. This segmentation allows the system to capture alignment information at different scales, with finer levels recovering details that may have been lost in coarser levels, thereby maintaining detection accuracy while still benefiting from the speed advantages of downsampling.
Solution Approach 2:
The patent performs preliminary alignment on downsampled image frames before processing higher-resolution frames. This preliminary action on lower-resolution data provides initial alignment parameters that guide subsequent processing of higher-resolution frames. This approach allows the system to benefit from the speed of downsampling while ensuring that final alignment accuracy is maintained through progressive refinement at higher resolutions.
Data Source
AI summary
Described examples relate to an apparatus comprising a memory for storing image frames and at least one processor. The at least one processor may be configured to receive a plurality of image frames from an image capture device and downsize each of the plurality image frames to generate a plurality of versions of each image frame at a plurality of different sizes. The at least one processor may also be configured to determine alignment information for a first version of a first image frame. The alignment information may include a first alignment vector for identifying image data in a first version of a second image frame that corresponds to image data in the first version of the first image frame. Further, the at least one processor may be configured to determine a first initial alignment vector for identifying image data in a first version of a third image frame based on at least the first alignment vector.