Tile-Based Moving Image Compression for Random Access
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing technologies face challenges in efficiently displaying high-definition moving pictures with varying angles of view, as they often require decoding and memory deployment of entire frames, leading to high arithmetic and memory costs, especially when only a portion of the image is displayed or enlarged.
Innovation Solution
A moving picture compression method that divides image frames into tile images, compresses them using spatial and temporal redundancy, and generates compression data with a pallet and index system, allowing for random access and reduced memory and processing costs by decoding only necessary regions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If compression rate is increased to handle large data size, then data storage efficiency is improved, but arithmetic operation cost upon decoding increases and granularity of data access increases
Solution Approach 1:
The image frame is divided into multiple tile images (e.g., 8x8 blocks), allowing the system to process and access only specific regions rather than entire frames. This segmentation enables localized decoding operations, reducing the arithmetic operation cost when only portions of the image need to be displayed or accessed.
Solution Approach 2:
The patent applies different compression strategies to different regions based on their importance and characteristics. By organizing data into tiles with local pallete tables, the system optimizes compression for each region independently, allowing high compression rates for less important areas while maintaining quality and enabling fast access for important regions.
2Quantity of substance
If compression rate is increased to handle large data size, then data storage efficiency is improved, but memory cost increases due to need to deploy overall area
Solution Approach 1:
By dividing the image into tile images with independent data structures, the system can load only the specific tiles needed for display into memory rather than deploying entire frames. This reduces memory requirements proportionally to the fraction of the image being displayed.
Solution Approach 2:
The patent extracts only the necessary tile images from the compressed data stream based on display requirements. The system can selectively decode and load specific tiles into memory, extracting exactly what is needed rather than loading complete frame data, thus reducing memory cost.
3Adaptability or versatility
If entire frame is decoded to display restricted region, then display flexibility is improved, but arithmetic operation cost increases
Solution Approach 1:
The tile-based structure allows the system to decode and display arbitrary combinations of tiles to create restricted regions or enlarged views. Since tiles are independently encoded with local pallete tables, the system can selectively decode only the tiles within the desired display region, maintaining full display flexibility while avoiding unnecessary decoding of other areas, thus reducing arithmetic operation cost.
4Adaptability or versatility
If entire frame is decoded to display restricted region, then display flexibility is improved, but processing time increases
Solution Approach 1:
By organizing image data into independently encoded tiles, the system can process and display only the specific tiles needed for the current view. This eliminates the need to decode entire frames, reducing processing time proportionally to the fraction of tiles being displayed while maintaining the ability to display any region or combination of regions.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A frame sequence of moving picture data is divided into a tile image sequence 250, and the color space of the tile image sequence 250 is converted to generate a YCbCr image sequence 252 (S10). Each frame is reduced to 1/2 time in the vertical and horizontal directions (S12), and a compression process is carried out to generate compression data 260 of a reference image (S14). The compression data 260 of the reference image is decoded and decompressed similarly as upon image display to restore a YCbCr image as the reference image, and a difference image sequence 262 is generated from the reference image and the original YCbCr image 252 (S16). Then, compression data 266 of a difference image is generated (S18), and compression data 268 obtained by connecting the compression data 260 of the reference image and the compression data 266 of the difference image is generated for every four frames of a tile image (S20).