Forward Discrete Cosine Transform Engine Pipelined Stages
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for discrete cosine transform (DCT) in JPEG encoding are resource-intensive and require substantial computation, making them inefficient for real-time processing in mobile devices with limited hardware resources.
Innovation Solution
A modified DCT engine that computes DCT in five stages using three coefficients and scales outputs with a plurality of scaling coefficients, distributing multiply operations across computation stages for a pipelined implementation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If conventional FDCT is implemented using standard algorithms, then image compression can be achieved, but hardware resources and computational complexity increase exponentially
Solution Approach 1:
The patent divides the DCT computation into five distinct stages (stages 1-5), where each stage performs specific transform operations on different parts of the data. This segmentation allows the complex 2D transform to be broken down into manageable computational blocks that can be processed efficiently with reduced hardware resources, directly resolving the contradiction between compression quality and hardware complexity.
Solution Approach 2:
The patent implements a dynamic computation schedule where multiply operations are distributed across the five stages rather than performed simultaneously. This dynamic approach allows the system to perform the same DCT transformation with fewer hardware resources by time-multiplexing operations, reducing the exponential hardware requirement while maintaining compression effectiveness.
2Measurement precision
If FDCT computation is performed with high precision, then encoding accuracy is maintained, but processing time and power consumption increase
Solution Approach 1:
By segmenting the DCT into five stages with distributed multiply operations, the patent enables precision encoding without requiring all multiplication operations to execute simultaneously. This temporal segmentation reduces peak power consumption while maintaining encoding accuracy through systematic computation progression.
Solution Approach 2:
The patent maintains continuous computation flow through the five stages, where output of one stage feeds into the next. This continuous action pattern ensures encoding accuracy is maintained while distributing power consumption over time rather than concentrating it in simultaneous operations, reducing overall power usage.
3Productivity
If DCT is implemented for real-time processing, then processing speed improves, but computational resource requirements increase
Solution Approach 1:
The five-stage segmentation enables real-time processing by allowing each stage to complete before the next begins, creating a pipeline that processes data continuously. This segmented approach achieves real-time productivity without requiring the exponential computational resources that would be needed for simultaneous full DCT computation.
Solution Approach 2:
The dynamic distribution of multiply operations across five stages enables real-time processing by time-multiplexing computations. This dynamic schedule allows the system to achieve high processing speed through sequential stage execution rather than requiring parallel computational resources, resolving the contradiction between productivity and device complexity.
Data Source
AI summary
Systems and methods are disclosed to perform fast discrete cosine transform (DCT) by computing the DCT in five stages using three coefficients, and scaling the outputs using a plurality of scaling coefficients.