Group Convolution Image Style Conversion YUV Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Image processing systems using deep neural networks face limitations in flexibility and processing speed, necessitating an improvement in these areas.
Innovation Solution
The method involves acquiring luminance and chrominance components in a YUV space of an image, performing group convolution processing to extract content and style features, and then fusing these features with target style features to convert the image into a desired style, utilizing a group convolutional neural network to reduce complexity and enhance processing efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If deep neural networks are used for image processing, then image style conversion capability is improved, but processing speed and system flexibility deteriorate
Solution Approach 1:
The patent segments the image processing into separate luminance (Y) and chrominance (U, V) component processing paths. Group convolution is applied independently to different color channels, allowing parallel processing that improves speed while maintaining style conversion capability. This segmentation resolves the contradiction by enabling faster processing without sacrificing the neural network's adaptability.
2Adaptability or versatility
If deep neural networks are used for image processing, then image style conversion capability is improved, but device complexity deteriorates
Solution Approach 1:
The patent divides the complex neural network into separate processing streams for luminance and chrominance components. By applying group convolution that processes different color channels independently, the system reduces computational complexity and device requirements while preserving the ability to perform sophisticated style conversion.
Solution Approach 2:
The patent extracts and processes only the essential chrominance components (U, V) separately from the luminance component (Y). This extraction approach simplifies the device complexity by focusing computational resources on the color information that carries style characteristics, rather than processing the entire image data uniformly.
3Reliability
If conventional convolution processing is used on full image data, then processing completeness is maintained, but processing complexity and time increase
Solution Approach 1:
The patent extracts only the chrominance components (U, V) that are essential for style information, while processing the luminance component (Y) separately. This selective extraction maintains processing completeness for style conversion while significantly reducing the amount of data that requires complex convolution operations, thereby reducing processing time.
Solution Approach 2:
The patent applies convolution processing selectively to only the necessary chrominance channels rather than processing all image data with the same level of complexity. This partial action approach maintains sufficient processing completeness for style conversion while reducing overall processing time and computational burden.
Data Source
AI summary
The present disclosure relates to a method, an apparatus and a device for converting a style of an image, wherein the method comprises: acquiring a luminance component (Y) and chrominance components (U, V) in a YUV space of an image to be processed; performing a group convolution processing on the luminance component (Y) and the chrominance components (U, V) in the YUV space of the image to be processed to obtain content features and style features of the image to be processed; and performing a fusion processing on the content features, the style features and target style features of the image to be processed to convert the image to be processed into an image of a target style.