Hash-Based Skip Encoding for Video Semiconductor Devices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image encoding apparatuses, such as those using the H.264 standard, perform unnecessary processing and consume excessive resources by always encoding based on prediction, even when the same images are input successively, leading to inefficiencies in memory usage and power consumption.
Innovation Solution
Incorporating a hash generator and reference hash list within the semiconductor device to compare hash values of input images, allowing for skip encoding when identical images are detected, thereby reducing processing and memory requirements without the need for large frame memory or motion search circuits.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If encoding is always performed based on prediction using large frame memory and motion search circuits, then encoding accuracy is maintained, but device complexity and power consumption increase significantly
Solution Approach 1:
The patent extracts only the essential function of frame memory by using a hash value storage unit that stores compressed hash values instead of full frame data. This extraction reduces device complexity while maintaining the core functionality of detecting identical frames for skip encoding.
Solution Approach 2:
The patent uses hash values as copies or representations of actual frame data. Instead of storing and comparing full frame images, the system stores and compares compact hash value copies, which dramatically reduces memory requirements and device complexity while preserving the ability to identify identical frames.
2Measurement precision
If full frame memory is used to store reference images for comparison, then identical frames can be detected accurately, but memory capacity and power consumption increase
Solution Approach 1:
The patent replaces full frame image storage with hash value storage. Each frame is represented by a compact hash value (e.g., 32-bit or 64-bit) instead of the original large image data, reducing memory capacity requirements by several orders of magnitude while maintaining comparison accuracy through hash matching.
Solution Approach 2:
The patent transforms the representation parameter of frame data from full pixel values to compressed hash values. This parameter change enables efficient storage and comparison by converting large-volume image data into small-volume fingerprint data that preserves the ability to identify identical frames.
3Productivity
If motion search circuits and large frame memory are implemented, then inter-frame prediction encoding is efficient, but busload and power consumption increase
Solution Approach 1:
The patent removes the need for complex motion search circuits by using hash value comparison instead. The system extracts only the essential comparison function and implements it through simple hash matching, eliminating energy-intensive motion estimation processes while maintaining encoding efficiency for identical frames.
Solution Approach 2:
The patent implements skip encoding that allows the encoder to rapidly skip detailed processing steps when identical frames are detected through hash comparison. This skipping mechanism rushes through unnecessary encoding operations for repeated frames, reducing power consumption and busload while maintaining high encoding efficiency.
Data Source
AI summary
A semiconductor device includes a hash generator, a reference hash list, a frame mode determination unit, and an intra prediction unit. The hash generator generates a hash value of a target frame to be encoded. The reference hash list is to record the hash value generated by the hash generator. The frame mode determination unit compares the hash value generated by the hash generator and the hash value in the reference hash list. The intra prediction unit performs intra prediction for the target frame to be encoded. When the hash value of the target frame to be encoded coincides with any of the hash values in the reference hash list, the intra prediction unit skips an encoding process, and outputs encoded information corresponding to any of the hash values in the reference hash list.