Portrait Image Processing via Frequency Component Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing dermabrasion and speckle removal algorithms for portrait images often result in significant loss of details, leading to unnatural and blunt effects, failing to maintain the clarity and realism of skin texture and portrait contours.
Innovation Solution
A method that processes portrait images by separating skin regions into high-frequency and low-frequency components, performing specific dermabrasion and speckle removal processes on each while protecting the portrait structure, and overlaying the results to retain realistic skin texture and maintain portrait clarity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional dermabrasion and speckle removal algorithms are applied to portrait skin regions, then blemishes are removed, but significant loss of skin texture details occurs resulting in unnatural and blunt effects
Solution Approach 1:
The patent segments the skin region image into multiple frequency components (low-frequency component image and high-frequency component image) using Gaussian filtering. The low-frequency component contains skin tone and basic texture information, while the high-frequency component contains detailed skin texture and blemish information. By processing different frequency components separately, the patent removes blemishes while preserving skin texture details, resolving the contradiction between blemish removal effectiveness and skin texture detail preservation.
2Reliability
If aggressive dermabrasion processing is applied to remove blemishes, then speckle removal effectiveness improves, but skin texture realism deteriorates
Solution Approach 1:
The patent applies different processing strategies to different regions and frequency components. In the high-frequency component image, speckle regions are identified and processed with speckle removal algorithms, while skin flat regions are processed with dermabrasion algorithms. The processed high-frequency component is then combined with the low-frequency component to produce the final output. This localized quality approach ensures speckle removal effectiveness while maintaining skin texture realism.
3Manufacturing precision
If frequency-based processing is applied to preserve skin texture, then skin texture realism is maintained, but processing complexity increases
Solution Approach 1:
The patent divides the processing into distinct frequency components using Gaussian filtering, which is a well-established and computationally efficient operation. The low-frequency component is obtained by Gaussian filtering, and the high-frequency component is obtained by subtracting the low-frequency component from the original image. This segmentation approach maintains skin texture realism while keeping the processing complexity manageable through the use of standard image processing operations.
Data Source
AI summary
The present application relates to a method and a device for processing a portrait image. The method includes determining a skin region, a skin flat region, a speckle region, and a portrait structure region. Performing a dermabrasion processing on the skin flat region and a speckle removal processing on the speckle region and overlaying the high-frequency output image after dermabrasion and speckle removal with the low-frequency output image after dermabrasion and speckle removal, to obtain a skin region output image. The method enables to present realistic skin texture and protect the clarity of portrait structure while achieving dermabrasion and speckle removal on skin in a portrait image.