Frequency Coefficient Selection for Image Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image compression techniques, such as those using discrete cosine transform (DCT), are inadequate for low latency applications like virtual or augmented reality due to limited compression efficiency and the introduction of artifacts like banding, which affect image quality.
Innovation Solution
A method that selectively encodes frequency coefficients based on their magnitude, using a transformation like 2D DCT, and generates an index to optimize compression, allowing differential compression across image areas and adapting to the viewer's gaze and device configuration, thereby prioritizing higher magnitude coefficients for better image quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If traditional discrete cosine transform compression is applied to reduce data volume, then bandwidth requirements are reduced, but compression efficiency is limited and processing time increases
Solution Approach 1:
The patent extracts only the most significant frequency coefficients (those with magnitude above a threshold) from the full set of DCT coefficients, discarding less important coefficients. This selective extraction achieves higher compression ratios by transmitting fewer coefficients while maintaining image quality, directly addressing the contradiction between data volume reduction and processing efficiency.
Solution Approach 2:
The patent applies different compression strategies to different frequency components based on their importance. High-magnitude coefficients that contribute most to image quality are preserved with higher fidelity, while low-magnitude coefficients are discarded or heavily compressed. This local differentiation optimizes both compression efficiency and processing speed.
2Manufacturing precision
If traditional discrete cosine transform compression retains lower frequency components, then image quality is maintained, but artifacts like banding occur
Solution Approach 1:
The patent changes the selection criterion from frequency-based (retaining only low frequencies) to magnitude-based (retaining coefficients with magnitude above threshold). This parameter change allows selective preservation of important high-frequency components while discarding insignificant low-frequency coefficients, eliminating banding artifacts while maintaining image quality.
Solution Approach 2:
Instead of the traditional approach of retaining low-frequency components and discarding high-frequency ones, the patent inverts the logic by retaining coefficients based on their magnitude regardless of frequency. This inversion allows important details (high-magnitude high-frequency coefficients) to be preserved while eliminating the banding problem caused by indiscriminate low-frequency retention.
3Manufacturing precision
If high resolution images are generated to appear lifelike, then image quality improves, but processing hardware requirements increase
Solution Approach 1:
The patent extracts only the essential frequency information needed to reconstruct high-resolution images, discarding redundant data. This extraction allows high-quality image generation with reduced computational burden, as the decoder only needs to process and reconstruct from the significant coefficients rather than handling full-resolution pixel data.
Solution Approach 2:
The patent applies partial action by processing only the necessary frequency coefficients rather than all pixel data. By focusing computational resources on the most significant coefficients that contribute most to perceived image quality, the system achieves high-resolution output with reduced hardware requirements.
4Loss of time
If compression is applied to reduce latency in virtual reality applications, then response time improves, but image artifacts increase
Solution Approach 1:
The patent changes the compression parameter from frequency-based selection to magnitude-based selection, allowing faster compression by simply comparing coefficient magnitudes against a threshold rather than performing complex frequency analysis. This parameter change reduces processing time (latency) while maintaining image quality by preserving significant coefficients.
Solution Approach 2:
The patent replaces complex traditional compression mechanisms with a simpler threshold-based filtering approach. Instead of elaborate quantization and encoding schemes, the system uses straightforward magnitude comparison and selective retention, significantly reducing processing time and latency while maintaining acceptable image quality.
Data Source
AI summary
A method of compressing image data representing one or more images, the method including obtaining pixel data from the image data, the pixel data representing an pixel array within the one or more images, applying a transformation to the pixel data to determine a set of frequency coefficients indicative of frequency coefficients of the pixel array, encoding a selected subset of the set of frequency coefficients, generating an index indicative of the encoded frequency coefficients and generating compressed image data using the encoded frequency coefficients and the index.