Adaptive DCT Filtering for Dark-Area Noise and Image Detail
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing noise filtering technologies struggle to effectively manage non-homogeneous noise in digital images, particularly in low light conditions, where dark area noise limits image brightening and conventional filters either remove too much detail in bright areas or ignore dark area noise.
Innovation Solution
An adaptive discrete cosine transform (DCT) noise filtering method that classifies image patches into bright and dark areas, applying separate noise suppression techniques to manage noise differently in each region, preserving texture and edge content.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If conventional noise filtering is applied to the entire image, then dark area noise is reduced, but bright area details are lost
Solution Approach 1:
The image is divided into multiple patches, and each patch is classified into bright area or dark area based on its average intensity. Different noise filtering strategies are then applied to different patch types, allowing aggressive denoising in dark areas while preserving details in bright areas.
Solution Approach 2:
The patent applies different noise suppression strengths to different spatial regions of the image. Dark area patches receive stronger noise filtering while bright area patches receive weaker or no filtering, creating a non-uniform filtering pattern that adapts to local image characteristics.
2Object-affected harmful factors
If strong noise filtering is applied to dark areas, then noise is suppressed, but image contrast is reduced
Solution Approach 1:
The patent dynamically adjusts the noise filtering strength based on the classification of each patch. By selecting different suppression levels for different patch types, the system adapts the filtering intensity to match the local noise characteristics and preserve image contrast where needed.
3Device complexity
If uniform filtering parameters are used for the entire image, then processing is simple, but noise management is ineffective in non-homogeneous images
Solution Approach 1:
Before applying noise filtering, the patent performs a preliminary classification of image patches into bright and dark areas based on their average intensity. This preliminary action enables the system to select appropriate filtering parameters for each region, effectively managing non-homogeneous noise while maintaining reasonable process complexity through automated classification.
Data Source
AI summary
A method includes dividing an image into overlapping image patches each having a specified size. The method also includes analyzing content of each image patch using a mathematical transform technique to classify each image patch into at least one class. The method further includes filtering each image patch for noise suppression by suppressing one or more transform coefficients of the image patch. An amount of suppression for each of the one or more transform coefficients is selected according to the at least one class of the image patch. In addition, the method includes reconstructing the filtered image patches into an output image.