Dual-Path Image Processing for Noise Reduction and Detail Preservation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Noise reduction in digital images captured in high ISO mode often results in blurred textures and gradient changes, leading to an unnatural appearance, as existing methods either blur details or fail to effectively reduce noise across different image areas.
Innovation Solution
A dual-path image processing method that captures a source image in Bayer format, performs first- and second-order image processes to produce noise-reduced and detailed images respectively, and blends these images with a noise map to maintain image detail and reduce noise, while avoiding blocky effects and discontinuities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If noise-reducing process is performed on high ISO captured images, then noise level is reduced, but image details and textures become blurred
Solution Approach 1:
The patent segments the image processing into multiple paths: a first-order image process that performs noise reduction and color interpolation, and a second-order image process that performs color interpolation without noise reduction. This segmentation allows different parts of the processing pipeline to handle different aspects of image quality independently.
Solution Approach 2:
The patent merges the outputs of the first-order and second-order image processes by blending their luminance images. The blended image combines the noise-reduced characteristics of the first-order process with the detail-preserving characteristics of the second-order process, achieving both noise reduction and detail preservation.
2Manufacturing precision
If sharpness algorithm is applied to blurry images, then edge sharpness is improved, but natural appearance is lost
Solution Approach 1:
The patent applies different processing qualities to different aspects of the image. The luminance channel receives noise reduction and blending operations to preserve natural appearance, while the chrominance channel maintains color information. This local differentiation of processing quality preserves natural appearance while improving edge sharpness where needed.
3Object-affected harmful factors
If different noise-reducing parameters are assigned for different areas, then noise reduction effectiveness is improved, but image uniformity deteriorates
Solution Approach 1:
The patent uses dynamic blending parameters that adapt to image characteristics. The blending weight is determined based on image features such as edge detection and texture analysis, allowing the system to dynamically adjust the mix between first-order and second-order process outputs for different regions of the image.
Data Source
AI summary
An image processing method is provided and includes the following steps. A source image arranged in a Bayer pattern is captured. A first-order image process is performed on the source image to produce a first image in YCbCr format. A second-order image process is performed on the source image to produce a second image in YCbCr format. Then, a noise-reducing process is performed on the first image in YCbCr format to produce a noise-reduced image. A blending operation is performed on luminance image of the noise-reduced image and luminance image of the second image in YCbCr format, and thereafter a chrominance image is combined with the blended image to produce a processed image. A noise reducing degree of the noise-reduced image is higher than that of the second image in YCbCr format.