GPU Texture Engine High Order Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing graphics processing units (GPUs) face challenges in performing high-order filtering due to hardware performance and power limitations, especially in embedded devices with limited computational power and memory capacity, which requires efficient image processing techniques for high-quality and high-resolution digital images.
Innovation Solution
The implementation of low-cost high-order filtering (LCHOF) techniques in a GPU using a modified texture engine that supports high-order filtering through a single shader instruction, fetching all necessary pixels from the local cache and blending them with pre-loaded weights, minimizing shader resource usage and memory pressure.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by moving object
If conventional filtering techniques are used in embedded devices with limited computational power and memory capacity, then hardware performance and power consumption are reduced, but filtering quality and processing capability deteriorate
Solution Approach 1:
The filtering operation is divided into two distinct phases: a compute-intensive phase where filter coefficients are calculated and stored in constant memory, and a data-intensive phase where neighboring pixels are fetched and processed. This segmentation allows the GPU to optimize each phase differently, reducing overall power consumption while maintaining filtering capability.
Solution Approach 2:
Filter coefficients are pre-calculated and stored in constant memory before the actual filtering operation. This preliminary action eliminates the need to recalculate filter coefficients during the filtering process, significantly reducing computational power requirements and energy consumption in embedded devices.
2Manufacturing precision
If high-order filtering is implemented with large filtering kernels, then image processing quality is improved, but hardware complexity and memory requirements increase
Solution Approach 1:
The texture engine is modified to perform multiple functions: it handles both the fetch operation for neighboring pixels and the blend operation for applying filter coefficients. This multi-functionality reduces hardware complexity by eliminating the need for separate dedicated filtering hardware while maintaining support for large filtering kernels and high image processing quality.
Solution Approach 2:
A modified texture engine acts as an intermediary between the standard GPU pipeline and the high-order filtering operation. This intermediary component is specifically designed to handle the unique requirements of HOF (fetching multiple pixels and blending with coefficients) while leveraging existing GPU architecture, thereby reducing overall hardware complexity.
3Measurement precision
If multiple shader instructions are used for high-order filtering, then filtering accuracy is improved, but shader resource usage and processing time increase
Solution Approach 1:
The fetch operation for neighboring pixels and the blend operation for applying filter coefficients are merged into a single unified operation within the modified texture engine. This consolidation reduces the number of shader instructions from multiple separate calls to a single instruction, significantly decreasing processing time while maintaining filtering accuracy through precise control of the blend operation.
Data Source
AI summary
This disclosure describes techniques for performing high order filtering in a graphics processing unit (GPU). In examples of the disclosure, high order filtering may be implemented on a modified texture engine of a GPU using a single shader instruction. The modified texture engine may be configured to fetch all source pixels needed for the high order filtering and blend them together with pre-loaded filtering weights.