An industrial image defocus blur suppression method, device, equipment and storage medium

CN122714271BActive Publication Date: 2026-10-16HANGZHOU ANMAISHENG INTELLIGENT TECH CO LTD
2 Cites 0 Cited by

Patent Information

Application Number
CN202611207603.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2026-08-11
Publication Date
2026-10-16
Estimated Expiration
2046-08-11

AI Technical Summary

Technical Problem

首先,传统卷积神经网络因其固有的局部感受野限制,难以建模大范围、空间变化的模糊,复原效果有限且易产生伪影

Benefits of technology

[0015]As can be seen, this invention inputs an industrial image into a defocus blur suppression model. The encoder of the defocus blur suppression model includes multiple sparse iterative deformable scanning modules connected in sequence. The sparse iterative deformable scanning module includes a sparse routing branch, an efficient bypass branch, and an iterative deformable scanning branch. Pixel annotations are added to the input feature map through the sparse routing branch. The pixel annotations include clear pixels and blurred pixels. The clear pixels in the input feature map are processed by the efficient bypass branch to obtain a first feature through a first sequence scan calculation and reset. The blurred pixels are corrected in coordinate and order through the first scan calculation of the iterative deformable scanning branch. After correction, the new pixel values ​​corresponding to the blurred pixels are arranged in order into a one-dimensional sequence. The one-dimensional sequence is processed by a second sequence scan calculation and reset to obtain a second initial feature. The second initial feature is fused with the blurred pixels in the input feature map to obtain a fused feature map. The second feature of the fused feature map is extracted through the second scan calculation of the iterative deformable scanning branch. The steps of the first scan calculation and the second scan calculation are the same. The first feature and the second feature are recombined to obtain the recombined feature output by the sparse iterative deformable scanning module. The recombined feature output by the encoder is input into the decoder of the defocus blur suppression model to obtain a defocus blur suppressed image.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The application discloses an industrial image defocus blur suppression method, device and equipment and a storage medium, and is applied to the field of image noise reduction. The method proposes an adaptive computing routing mechanism, distinguishes clear pixels and blurred pixels in an image, and dynamically allocates computing resources to only the blurred pixels, so that the operation efficiency is greatly improved without sacrificing the accuracy. An iterative scanning refinement strategy is designed. For the features routed to the blurred pixels, an iterative deformable scanning mechanism from coarse to fine is introduced. Through twice offset prediction and feedback, it can penetrate through serious blur, gradually approach and align to the real edges and texture structures of the object, and realize high-fidelity detail recovery. Through the adaptive computing routing mechanism and the iterative scanning refinement strategy, the application realizes high-definition, high-efficiency and high-robustness industrial image defocus blur suppression.
Need to check novelty before this filing date? Find Prior Art