Image Data Compression for Lower SRAM and DRAM Bandwidth
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing systems face challenges in efficiently managing SRAM and DRAM resources, particularly in high-throughput applications, due to high bandwidth requirements and power consumption, necessitating the reduction of SRAM storage and optimizing DRAM bandwidth usage.
Innovation Solution
A method of compressing image data by determining representative values based on relationships between neighboring data positions, reducing the bit depth of each data value while preserving the number of data positions, thereby minimizing SRAM requirements and optimizing DRAM bandwidth.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If SRAM is used to store image data for high throughput processing, then processing speed and latency are improved, but SRAM storage requirements and cost increase
Solution Approach 1:
The patent segments image data into multiple sequences (e.g., by color channels or spatial regions) and processes each sequence independently with its own compression parameters. This allows selective compression of different data portions, reducing overall SRAM storage requirements while maintaining processing throughput by parallelizing sequence handling.
Solution Approach 2:
The patent dynamically adjusts compression parameters (such as bit depth reduction from 12-bit to 8-bit, or compression ratios) based on image content characteristics and processing requirements. By changing these parameters adaptively, the system optimizes the balance between processing speed and SRAM storage usage, using higher compression for less critical data and lower compression for important regions.
2Productivity
If DRAM bandwidth is increased to handle high throughput applications, then data transfer capacity is improved, but power consumption and bandwidth availability for other components worsen
Solution Approach 1:
The patent applies compression to image data before it is written to DRAM, reducing the total volume of data that needs to be transferred. This preliminary compression action decreases DRAM bandwidth requirements and associated power consumption, while still providing sufficient data throughput for high-performance processing by optimizing the compressed data flow.
3Quantity of substance
If DRAM is used instead of SRAM for image data storage, then cost per bit is reduced, but latency and bandwidth availability worsen
Solution Approach 1:
The patent divides image data into multiple sequences that can be processed and compressed independently. This segmentation allows the system to use smaller, more frequent DRAM write operations rather than large bulk transfers, reducing the impact of DRAM latency and improving time-critical processing while still benefiting from lower storage costs compared to SRAM.
Data Source
AI summary
Image data includes a plurality of data values arranged in two or more sequences, each sequence comprising a first number of data positions each having a data value of the plurality, and each data value comprising a second number of bits. Upon receiving each sequence of the two or more sequences, the data values of the sequence are processed to determine, for each given data position of the first number of data positions, a representative value for the given data position based on a relationship between the data value at the given data position and the data value at one or more neighbouring data positions.