Distributed Image Coding With Side-Information Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image compression techniques require significant computational resources and power, making them unsuitable for resource-constrained environments like satellite imaging, where encoders need to be lightweight and decoders can utilize more resources.
Innovation Solution
A low-complexity encoding approach that shifts complexity to decoding, using syndrome coding and distributed source coding to compressively measure signals, exploiting structural correlations among bands, and utilizing side information for accurate prediction and reconstruction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional transform coding methods (JPEG, JPEG2000) are used for image compression, then compression efficiency is improved, but encoder computational complexity and power consumption increase significantly
Solution Approach 1:
The patent inverts the conventional compression paradigm by placing computational complexity in the decoder rather than the encoder. The encoder performs simple compressive sensing measurements, while the decoder handles the complex iterative reconstruction algorithms, thereby resolving the contradiction between compression efficiency and encoder complexity for resource-constrained platforms.
Solution Approach 2:
The patent replaces traditional mechanical transform coding operations (DCT, wavelets) with compressive sensing linear measurements followed by iterative optimization-based reconstruction. This substitution enables efficient encoding while maintaining compression performance, as the heavy computational burden is shifted to the decoder side.
2Device complexity
If distributed coding approaches are used to reduce encoder complexity, then encoder computational resources are reduced, but the ability to exploit signal structure and accurately estimate bit rate deteriorates
Solution Approach 1:
The patent applies preliminary action by performing compressive sensing measurements at the encoder before transmission. These measurements capture the essential signal structure in a compressed form, enabling the decoder to exploit correlations and reconstruct the signal accurately without requiring complex encoder operations.
Solution Approach 2:
The patent introduces compressive sensing measurements as an intermediary between the original signal and the transmitted data. These measurements serve as a bridge that preserves signal structure information while reducing the data volume, allowing distributed coding to maintain structure exploitation capability without increasing encoder complexity.
3Device complexity
If compressive sensing measurements are used at the encoder, then encoder complexity is reduced, but decoder computational requirements increase
Solution Approach 1:
The patent changes the operational parameters by using random projection matrices for compressive sensing measurements at the encoder, which are computationally simple. At the decoder, iterative optimization algorithms with adjustable regularization parameters are employed, allowing the system to balance decoder computational power requirements while maintaining reconstruction quality.
Data Source
AI summary
A method for reconstructing an uncompressed signal. The method includes obtaining an encoded signal corresponding to the signal. Obtaining side information about the signal and using the side information to obtain a prediction of dithered linear measurements of the signal. Using the prediction of the dithered linear measurements and encoded quantized dithered linear measurements to obtain quantized linear measurements of the signal based on processing each bitplane iteratively, starting from a least significant level bitplane to a most significant level bitplane. At each iteration, a prediction of each bitplane is formed using the prediction of the dithered linear measurements and the bitplanes processed in the previous iterations. Wherein each code for each bitplane is used to correct each bitplane prediction. Reconstructing the signal as a reconstructed signal using the recovered quantized dithered linear measurements, wherein the steps are performed in a processor of a decoder.