Line-Based Image Compression Using Spatial Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Digital video products face challenges in reducing memory bandwidth requirements, especially at high video resolutions like D1 (720×480) or higher, due to limited memory bandwidth caused by cost and power constraints, which affects real-time video data transfer and processing.
Innovation Solution
The implementation of a method for compressing and decompressing digital image data using spatial prediction and entropy encoding, where pixel predictors are computed using neighboring pixels, and variable length codes are selected based on coding selection criteria, reducing the need for explicit signaling in the compressed data and optimizing memory usage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If video data is transferred in real-time from preview engine to external memory and to/from external memory to video encoder, then video processing functionality is achieved, but memory bandwidth requirements become excessively high
Solution Approach 1:
The video data transfer process is segmented into multiple stages: compression in preview engine, selective storage in external memory, decompression when needed, and encoding. This segmentation allows only compressed data to be stored and transferred to external memory, reducing the memory bandwidth requirement while maintaining real-time processing capability.
Solution Approach 2:
Video data is compressed in advance in the preview engine before being stored in external memory. This preliminary compression action reduces the data volume that needs to be transferred to and from external memory, thereby reducing memory bandwidth requirements while enabling real-time processing when data is retrieved and decompressed.
2Measurement precision
If video resolution is increased to D1 (720×480) or higher, then picture quality is improved, but memory bandwidth requirements increase significantly
Solution Approach 1:
The data representation parameter is changed from uncompressed raw video data to compressed video data. This parameter change reduces the data volume by a factor of 2 or more, allowing high-resolution video (D1 or higher) to be processed and stored with reduced memory bandwidth requirements while maintaining picture quality.
3Quantity of substance
If compressed video data is stored in external memory, then memory bandwidth requirements are reduced, but decompression complexity is increased
Solution Approach 1:
The compression algorithm used in the preview engine is designed to be reversible with a corresponding decompression algorithm. The decompression process uses the same structural patterns and prediction methods as the compression, allowing the system to self-serve by using identical algorithmic approaches for both compression and decompression, thereby managing complexity efficiently.
4Quantity of substance
If spatial prediction with neighboring pixels is used for compression, then compression ratio is improved, but computational complexity increases
Solution Approach 1:
Spatial prediction uses only locally adjacent pixels (neighboring pixels) to predict the current pixel value, rather than using global image information. This local approach achieves good compression ratios by exploiting local correlations while keeping computational complexity low, as it only requires accessing and processing a small number of neighboring pixel values.
Data Source
AI summary
A method of compressing digital image data is provided that includes selecting an entropy code for encoding a line of pixels in the digital image data, wherein the entropy code is selected from a plurality of variable length entropy codes, using spatial prediction to compute a pixel predictor and a pixel residual for a pixel in the line of pixels, and selectively encoding the pixel residual using one of the entropy code or run mode encoding.