Fractional Interpolation via Integer Downsampling Sequences
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing interpolation techniques, particularly in autonomous or semi-autonomous vehicles, are inflexible and require complex algorithms or specialized hardware to perform fractional interpolation, which is necessary for adjusting image resolutions from diverse image sensors.
Innovation Solution
The implementation of efficient computational techniques for fractional interpolation using an interpolation sequence that combines multiple rounds of interpolation and downsampling, specifically designed for convolutional engines or processors, allowing for flexible adjustment of image resolutions without the need for complex hardware controls.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If complex algorithms or specialized hardware are used for fractional interpolation, then interpolation precision is improved, but device complexity increases
Solution Approach 1:
The fractional interpolation process is segmented into multiple discrete steps: integer interpolation, downsampling, and fractional adjustment. Each step uses simpler operations that can be implemented with basic convolutional processor components, avoiding the need for complex specialized hardware while achieving precise fractional interpolation results
Solution Approach 2:
The patent introduces intermediate processing steps (integer interpolation followed by downsampling) as mediators between the input image and the final fractional interpolation output. These intermediate steps break down the complex fractional interpolation task into manageable operations that can be performed using standard convolutional processor operations
2Measurement precision
If complex algorithms are used for fractional interpolation, then interpolation precision is improved, but computational time increases
Solution Approach 1:
The patent performs preliminary integer interpolation and downsampling operations before executing the final fractional adjustment. By pre-processing the image with simpler operations, the computationally intensive fractional interpolation step operates on a smaller, pre-processed dataset, reducing overall computational time while maintaining precision
3Measurement precision
If specialized hardware is used for fractional interpolation, then interpolation precision is improved, but ease of operation deteriorates
Solution Approach 1:
The patent designs a fractional interpolation method that operates using standard convolutional processor operations, making the technique universally applicable to existing convolutional processors without requiring specialized hardware. This multi-functional approach allows the same hardware to perform both standard convolution operations and fractional interpolation, improving ease of operation while maintaining precision
Data Source
AI summary
Systems and methods for enhanced fractional interpolation for a convolutional processor in autonomous or semi-autonomous systems. An example method includes obtaining images from image sensors positioned about a vehicle. For a first image, an interpolation sequence to interpolate the first image according to a fractional interpolation value is determined, with the interpolation sequence including combinations of respective integer interpolations and integer downsamplings. The first image is interpolated according to the interpolation sequence.