Content-Adaptive DNN Deblocking for Block-Wise Image Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face challenges in efficiently reducing redundancy and distortion in block-wise image compression, particularly in intra prediction and motion compensation, leading to suboptimal compression ratios and quality.
Innovation Solution
Employing a deep neural network (DNN) for deblocking in video decoders, trained adaptively based on content, to improve deblocking at block boundaries, using deblocking parameters determined through a content-adaptive training process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If traditional intra coding and block-wise compression are used, then bandwidth and storage requirements are reduced, but distortion and visual quality deteriorate at block boundaries
Solution Approach 1:
A deep neural network (DNN) is introduced as an intermediary component between the block-wise compression process and the final reconstructed image. The DNN specifically targets block boundary regions and applies learned filtering operations to reduce artifacts while preserving important image features. This intermediary processing step resolves the contradiction by maintaining compression efficiency while improving visual quality at problematic boundary regions.
Solution Approach 2:
The deblocking filter applies different processing strategies to different regions of the image. Block boundary regions receive specialized DNN-based filtering to address artifacts, while other regions maintain standard compression processing. This local differentiation allows the system to improve quality where needed without compromising overall compression efficiency.
2Manufacturing precision
If deep neural network based deblocking is employed, then visual quality and compression efficiency improve, but computational complexity and processing time increase
Solution Approach 1:
The deblocking process is segmented into distinct operational phases: identification of block boundary regions, application of DNN-based filtering to these specific regions, and combination with the compressed image data. By segmenting the processing to focus computational resources only on problematic boundary regions rather than the entire image, the system achieves quality improvement while controlling overall computational complexity.
Solution Approach 2:
The DNN model is trained in advance on large datasets to learn optimal deblocking operations. This preliminary training phase allows the model to be deployed during actual compression/decoding with minimal real-time computational overhead, as the complex learning has already been performed during offline training.
3Reliability
If content-adaptive training is used, then deblocking performance is optimized for specific image types, but training time and data processing requirements increase
Solution Approach 1:
The system employs content-adaptive training that dynamically adjusts the DNN training process based on the characteristics of the input image data. Different image types (e.g., natural scenes, synthetic images, medical imagery) trigger different training configurations and parameter optimizations. This dynamic adaptation allows the system to achieve high reliability for specific content types while managing training time through selective and targeted training approaches.
Data Source
AI summary
Aspects of the disclosure provide a method, an apparatus, and non-transitory computer-readable storage medium for video decoding. The apparatus includes processing circuitry that reconstructs blocks of an image that is to be reconstructed from a coded video bitstream. The processing circuitry decodes first deblocking information in the coded video bitstream including a first deblocking parameter of a deep neural network (DNN) in a video decoder. The first deblocking parameter of the DNN is an updated parameter that has been previously determined by a content adaptive training process. The processing circuitry determines the DNN for a first boundary region comprising a subset of samples in the reconstructed blocks based on the first deblocking parameter included in the first deblocking information. The processing circuitry deblocks the first boundary region comprising the subset of samples in the reconstructed blocks based on the determined DNN corresponding to the first deblocking parameter.