Video Classification Engine for Perceptual Bit Allocation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional video encoders face challenges in rate control, leading to under-allocation of bits in areas with high motion or perceptual significance, resulting in noticeable artifacts, as they typically use constant quantization resolution or equal bits for every macroblock.
Innovation Solution
A system and method for video classification that allocates bits based on perceptual quality by using a classification engine to detect target objects like skin, adjusting quantization levels, and generating a quantization map to allocate more or fewer bits to blocks, ensuring optimal bit allocation and resolution according to the importance of video content.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a video encoder uses constant quantization resolution for every block, then the encoding process is simple and fast, but bits are under-allocated in areas that are well predicted and perceptually significant, degrading quality
Solution Approach 1:
The patent applies local quality by classifying video blocks into different types (e.g., skin, sky, water, tree) and applying different quantization parameters to different block types. This allows perceptually important regions like skin to receive finer quantization (more bits) while less important regions use coarser quantization (fewer bits), thereby improving overall video quality without uniformly increasing bit rate.
Solution Approach 2:
The patent implements dynamic quantization adjustment by using a classification engine that analyzes each video block's characteristics in real-time and dynamically selects appropriate quantization parameters. This dynamic approach replaces static constant quantization with adaptive per-block quantization, improving quality while maintaining encoding efficiency through automated classification.
2Device complexity
If a video encoder uses the same number of bits for every macroblock, then the encoding process is simple, but bits are under-allocated to complex areas (high motion) such that blocking artifacts become noticeable
Solution Approach 1:
The patent classifies macroblocks into different types based on local characteristics (motion complexity, content type) and applies different quantization strategies to each class. Complex areas with high motion are identified and allocated more bits through finer quantization, preventing blocking artifacts, while simple areas use coarser quantization to maintain overall bit rate efficiency.
Solution Approach 2:
The patent performs preliminary classification of each macroblock before encoding, identifying areas with high motion or complex content. This preliminary action allows the encoder to pre-determine appropriate quantization parameters for each block, ensuring that complex areas receive sufficient bits before the actual encoding process, thereby preventing blocking artifacts.
3Manufacturing precision
If rate control systems choose a quantization level to balance between perceptual significance and complexity, then video quality improves, but the system complexity increases
Solution Approach 1:
The patent segments the video content into classified blocks based on content type (skin, sky, water, tree, etc.) and motion characteristics. This segmentation allows the rate control system to manage different quantization strategies for different block types independently, simplifying the overall control logic while achieving perceptually optimized quality through category-based parameter selection.
Data Source
AI summary
Described herein is a method and system for video classification. The method and system can use predetermined color ranges to classify a video block. On a relative basis, a greater number of bits can be allocated to perceptually important video data such as skin. The quantization is adjusted accordingly. Determining relative quantization shifts for macroblocks in a picture prior to video encoding enables a corresponding bit allocation that can improve the tradeoff between perceptual quality and bit rate.