Image Domain Conversion With Feature Correlation and Content Consistency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing methods using machine learning models often result in significant content distortion during image domain conversion, leading to poor processing effects.
Innovation Solution
An image processing method utilizing a generative adversarial network (GAN) and cycle-consistent adversarial network (cycleGAN) to train a model that minimizes content changes and maintains image consistency by adjusting model parameters based on feature and distribution correlation degrees, employing an information bottleneck theory to ensure image content invariance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional image processing methods are used to process high-definition video streams, then processing speed can be maintained, but image quality deteriorates due to lossy compression artifacts
Solution Approach 1:
The patent divides the image processing task into multiple processing units that operate in parallel. Each processing unit handles specific image data blocks independently, enabling simultaneous processing of multiple regions. This segmentation approach maintains high processing speed while applying more sophisticated processing algorithms to improve image quality and reduce compression artifacts.
Solution Approach 2:
The patent introduces an image buffer as an intermediary component between the video stream input and the processing units. The buffer stores image data and coordinates the flow of data to multiple processing units, enabling efficient parallel processing without data loss. This intermediary mechanism ensures that high-definition image data is preserved while being processed at high speed.
2Reliability
If multiple image processing operations are performed sequentially, then processing thoroughness is maintained, but processing time increases
Solution Approach 1:
The patent segments the image processing operations into multiple independent processing units that can execute simultaneously. Each processing unit performs specific operations (e.g., noise reduction, sharpening, color correction) on different portions of the image data in parallel, maintaining comprehensive processing coverage while significantly reducing total processing time compared to sequential execution.
Solution Approach 2:
The patent performs preliminary preparation of image data by storing it in the image buffer before processing begins. This preliminary action organizes the data structure and makes it readily accessible to multiple processing units, enabling them to start processing immediately without waiting for data preparation, thus reducing overall processing time while maintaining thoroughness.
3Productivity
If image data is buffered and processed in parallel, then processing efficiency improves, but device complexity increases
Solution Approach 1:
The patent designs processing units with multi-functional capabilities that can handle various image processing operations. Each processing unit is configured to perform multiple types of processing tasks, reducing the total number of specialized components needed. This universal approach improves processing efficiency through parallel operation while limiting the increase in device complexity by reusing the same hardware resources for different functions.
Solution Approach 2:
The patent creates multiple copies of processing units that can operate simultaneously on different image data blocks. Rather than designing a single complex processing system, the solution replicates simpler processing unit designs in parallel. This copying strategy achieves high processing efficiency through concurrency while keeping each individual processing unit relatively simple, thus limiting overall device complexity.
Data Source
Figure 1~2
Figure 3~5
Figure 6
AI summary
The present application relates to an image processing method and apparatus, a computer device, a storage medium, and a program product. The method comprises: inputting a first image belonging to a first image domain into an image processing model to be trained, to obtain a second image belonging to a second image domain; acquiring an association degree between an image feature of the first image and an image feature of the second image to obtain a target feature association degree; acquiring an association degree between feature value distribution of image features of the first image and feature value distribution of image features of the second image; obtaining a distribution association degree; and adjusting model parameters of the image processing model in a direction in which the target feature association degree is increased and the distribution association degree is increased, to obtain a trained image processing model. By means of the method, the image processing effect can be improved. The image processing model in the present application may be an artificial intelligence-based neural network model, and an artificial intelligence cloud service is provided on the basis of the model.