Lossless Image Compression via Dynamic Quantization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Video compression and decompression processes consume significant DRAM bandwidth, potentially slowing down real-time operations due to frequent access during processing, compression, and motion compensation.
Innovation Solution
A system and method for lossless and near-lossless image compression using a virtual buffer to dynamically adjust quantization steps in video encoding and decoding, reducing the need to transmit quantization information and optimizing bandwidth usage by simulating a buffer that adjusts based on fullness or emptiness, thereby minimizing memory access.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If video compression and decompression processes are performed with frequent DRAM access during processing, compression, and motion compensation, then image quality and processing accuracy are maintained, but DRAM bandwidth consumption increases significantly
Solution Approach 1:
The patent applies preliminary action by pre-calculating and storing motion compensation data in on-chip memory before actual video processing occurs. The motion estimation results and compensation parameters are prepared in advance, eliminating the need for repeated DRAM accesses during the actual compression and decompression operations, thus reducing bandwidth consumption while maintaining processing accuracy
Solution Approach 2:
The patent introduces an intermediary on-chip memory buffer that acts as a mediator between the video processing units and the main DRAM. This intermediate storage layer holds frequently accessed motion compensation data, reducing direct DRAM bandwidth consumption while ensuring that high-quality image processing can proceed without memory access bottlenecks
2Productivity
If reconstructed frames are stored in DRAM for displaying and motion compensation operations, then real-time video processing is enabled, but DRAM bandwidth is consumed at high rates
Solution Approach 1:
The patent implements a nested memory architecture where on-chip memory is embedded within the video processing unit, which itself is part of the larger system connected to DRAM. This nested structure allows the innermost memory layer to handle frequent access to motion compensation data, reducing the need for outer DRAM accesses while maintaining real-time processing capability
Solution Approach 2:
The video processing unit performs self-service by maintaining its own local on-chip memory for storing motion compensation data and reconstructed frames. This self-contained memory system allows the processor to service its own data needs without constantly accessing external DRAM, enabling real-time processing while minimizing bandwidth consumption
Data Source
AI summary
Presented herein is lossless and near-lossless image compression. In one embodiment, there is described a method for encoding data. The method comprises receiving said data; quantizing a predetermined amount of the data with a predetermined quantization step; determining how many bits represent the predetermined amount of the data quantized with the predetermined quantization step; and wherein if the amount of data exceeds a predetermined amount of bits, changing the quantization step.