PReLU Neural Video Filtering Blocks for Lower Processing Complexity

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Neural network-based filtering techniques for video data are highly complex and require significant processing power, limiting their widespread application in video coding devices.

Innovation Solution

Implement simplified neural network processing blocks, including 1×1 convolutional filters, parametric rectified linear unit (PRELU) filters, and 3×3 convolutional filters, to enhance video data filtering while reducing processing requirements.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If neural network-based filtering techniques are applied to video data, then video filtering performance is improved, but processing complexity and processing power requirements increase

Engineering Contradiction:

Improvevideo filtering performanceVSAvoidprocessing complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The neural network filter is divided into multiple processing blocks (first neural network processing block, second neural network processing block, etc.), each handling specific filtering operations. This segmentation reduces the complexity burden on any single device by distributing the filtering task across multiple manageable blocks that can be implemented independently

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent employs parameter changes through PRELU (parametric rectified linear unit) filters that adaptively adjust filter coefficients based on input data characteristics. This allows the network to optimize filtering performance dynamically while maintaining lower computational complexity compared to fixed complex neural network architectures

Inventive Principle:

Principle #35Parameter changes

2Reliability

If neural network-based filtering techniques are applied to video data, then video filtering performance is improved, but processing power requirements increase

Engineering Contradiction:

Improvevideo filtering performanceVSAvoidprocessing power needs

Core Design Contradiction:

ReliabilityVSPower

Solution Approach 1:

By segmenting the filtering process into multiple blocks with specific functions (first block for initial filtering, second block for refinement), the patent reduces the processing power requirement at any single stage while maintaining overall filtering effectiveness through sequential processing

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent applies filtering operations selectively to different portions of video data based on complexity requirements. Not all video blocks require the same level of processing, allowing the system to apply full neural network filtering only where necessary while using simpler processing for other regions

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS20250324050A1Reduced complexity multi-mode neural network filtering of video data

Publication Date: 2025.10.16 QUALCOMM INC

AI summary

An example device for filtering video data includes a memory configured to store video data; and a processing system comprising one or more processors implemented in circuitry, the processing system being configured to: apply one or more neural network processing blocks to intermediate filtered video data, each of the neural network processing blocks including a first 1×1 convolutional filter, a parametric rectified linear unit (PReLU) filter, a second 1×1 convolutional filter, and a 3×3 convolutional filter; apply additional neural network processing blocks to output of the one or more neural network processing blocks to form filtered video data; and output the filtered video data.