Video Coding Normalization Signaling for Neural Network Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding standards like ITU-T H.264, ITU-T H.265, and ITU-T H.266 face challenges in efficiently signaling normalization techniques for video coding, particularly in handling intra and inter prediction, transform, and entropy coding, which can impact data compression efficiency and decoding accuracy.
Innovation Solution
Implementing a normalization technique determination and signaling mechanism in video coding systems, including methods for encoding and decoding video data, which involves determining a normalization technique from multiple options and signaling it in a bitstream, and performing normalization on decoded data for input into a neural network.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If normalization techniques are not signaled in the bitstream, then device complexity is reduced, but decoding accuracy deteriorates
Solution Approach 1:
The patent applies parameter changes by introducing normalization technique parameters (such as normalization type and shift amount) into the bitstream signaling mechanism. This allows the decoder to adjust its processing parameters based on the encoded normalization technique, thereby improving decoding accuracy without requiring complex additional hardware structures. The encoder signals these parameters using existing syntax element frameworks, changing the parameter space rather than the fundamental system architecture.
2Adaptability or versatility
If multiple normalization techniques are supported, then adaptability is improved, but device complexity increases
Solution Approach 1:
The patent implements dynamics by making the normalization technique selection adaptive rather than static. The encoder determines the appropriate normalization technique based on the specific video content characteristics (such as prediction mode, transform type, and residual data properties), and signals this selection in the bitstream. The decoder dynamically adjusts its processing based on the received normalization technique parameters, allowing the system to adapt to different video scenarios without requiring all possible normalization techniques to be simultaneously implemented in hardware.
Solution Approach 2:
The patent applies segmentation by dividing the normalization process into distinct technique options (such as different normalization types and shift amounts) that can be independently selected and signaled. Rather than implementing a single complex normalization unit that handles all cases, the system segments the normalization functionality into multiple selectable techniques, with the encoder choosing the most appropriate one for each coding unit and signaling it to the decoder.
3Productivity
If normalization is performed on decoded data, then data compression efficiency is improved, but processing time increases
Solution Approach 1:
The patent applies preliminary action by performing the normalization technique determination and parameter signaling during the encoding phase, before the actual decoding and reconstruction process. The encoder analyzes the video data characteristics and pre-determines the optimal normalization technique, then embeds this information in the bitstream. During decoding, the normalization process can be efficiently executed by simply applying the pre-determined parameters without requiring complex real-time analysis, thus reducing processing time while maintaining compression efficiency.
Data Source
AI summary
A device may be configured to receive a bitstream. The device may decode the bitstream according to a video coding technique. The device may determine a normalization technique from one or more normalization techniques. The device may perform the normalization technique on the decoded bitstream to generate feature data for input into a neural network. Normalizations techniques may include a combination of a linear normalization techniques, scaling down normalization techniques, and logarithmic normalization techniques.