Facial Image Denoising via Reference Texture Transfer
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image denoising techniques often result in facial images that appear too 'flat' and unrealistic due to the removal of noise, as they fail to effectively differentiate and preserve face-specific details.
Innovation Solution
A method that identifies noisy facial portions in images and corresponding cleaner portions from other images, determines filter parameters based on noise levels, and uses these parameters to smooth noise while adding face-specific details, employing techniques like bilateral filtering and non-parametric Markov random fields for texture optimization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If conventional denoising techniques are used to reduce image noise, then noise is reduced, but facial details become flat and unrealistic
Solution Approach 1:
The patent applies different processing to different regions of the image by segmenting facial regions from non-facial regions. Facial regions undergo a two-step process (smoothing followed by detail enhancement), while non-facial regions receive only smoothing. This local differentiation preserves facial details while reducing noise, resolving the contradiction between noise reduction and detail preservation.
Solution Approach 2:
The image processing is segmented into distinct stages: noise smoothing and detail enhancement. The facial region detection and segmentation enable separate processing pipelines. This segmentation allows the system to apply appropriate operations to different parts of the image, maintaining facial realism while reducing overall noise.
2Object-affected harmful factors
If noise smoothing is applied to the entire image, then noise is reduced, but face-specific details are lost
Solution Approach 1:
The patent performs smoothing as a preliminary action on the entire image first, then applies detail enhancement as a subsequent action on facial regions. This sequential approach ensures that noise is reduced globally while face-specific details are restored locally, preventing permanent loss of important information.
Solution Approach 2:
The patent extracts and isolates facial regions from the rest of the image using face detection and segmentation. By taking out the facial regions for specialized processing, the system can apply detail enhancement operations specifically where needed, preserving face-specific details that would otherwise be lost in global smoothing.
3Manufacturing precision
If multiple images are processed to find corresponding portions, then cleaner facial portions can be identified, but processing complexity increases
Solution Approach 1:
The patent uses multiple images of the same subject to create reference copies for comparison. By identifying corresponding portions across multiple images, the system can select the cleanest version of each facial region. This copying approach improves image quality without requiring complex single-image processing algorithms.
Data Source
AI summary
A method, computer program product, and computer system for identifying a first portion of a facial image in a first image, wherein the first portion includes noise. A corresponding portion of the facial image is identified in a second image, wherein the corresponding portion includes less noise than the first portion. One or more filter parameters of the first portion are determined based upon, at least in part, the first portion and the corresponding portion. At least a portion of the noise from the first portion is smoothed based upon, at least in part, the one or more filter parameters. At least a portion of face specific details from the corresponding portion is added to the first portion.