Single-exposure compressive imaging reconstruction method and system fusing pre-trained language model

By combining multimodal data alignment and sparse controlled diffusion Transformer, the island problem of single-exposure compressed imaging reconstruction algorithm is solved, realizing efficient unified mapping and high-precision reconstruction of cross-modal data, breaking the island barrier of traditional models, and possessing strong zero-shot adaptation capability.

CN122312380APending Publication Date: 2026-06-30CHENGDU INSTITUTE OF BIOLOGY CHINESE ACADEMY OF SCIENCES
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202610772811.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-06-01
Publication Date
2026-06-30

AI Technical Summary

Technical Problem

Existing single-exposure compressed imaging reconstruction algorithms suffer from problems such as high generation randomness, lack of physical realism, isolated models, loss of local details in feature extraction, lack of accurate semantic guidance, and insufficient generalization ability. This leads to difficulties in transferring prior knowledge between different imaging tasks and high system development costs.

Method used

The system employs a multimodal data alignment and physical observation module, a mask-gated feature extraction module, a multimodal large language model fusion module, and a dual-condition guided sparse controlled diffusion Transformer module. Combined with a lightweight spatiotemporal encoder and a pre-trained language model, it achieves unified mapping of cross-modal data and high-fidelity feature extraction. It utilizes the pre-trained large language model to generate high-dimensional semantic understanding and performs denoising and reconstruction through the sparse controlled diffusion Transformer.

Benefits of technology

It achieves a unified network architecture for various imaging tasks, is compatible with different compression ratios and modalities, preserves high-frequency physical features, has strong zero-sample generalization ability, outputs high-precision reconstruction results, and can adapt to new imaging systems without repeated training.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122312380A_ABST
    Figure CN122312380A_ABST
Patent Text Reader

Abstract

This invention specifically relates to a single-exposure compressed imaging reconstruction method and system that integrates pre-trained language models, solving the problem that existing technologies can only decode a single imaging modality. It employs a model framework of "physical constraints-semantic guidance-sparse controlled diffusion," uniformly mapping cross-modal compressed data to align the data stream. On one hand, it utilizes a lightweight mask-gated constraint feature extraction module to extract high-fidelity underlying physical features; on the other hand, it cascades a pre-trained multimodal large language model to generate a global semantic representation. In the controlled diffusion large model denoising and reconstruction stage, an asymmetric sparse controlled DiT architecture is used to solve the problem of easily losing high-frequency details, and computational redundancy is significantly reduced by decreasing the number of controllable modules and layers. Through this hardware-software co-design, 3D positional encoding mechanism, and text prompt word design, a single model can adaptively reconstruct high-precision images with multiple imaging modalities and compression rates.
Need to check novelty before this filing date? Find Prior Art