Adaptive Image Sensor Noise Reduction Kernel
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing digital camera technologies face challenges in effectively reducing noise generated by image sensors without damaging the original image quality or requiring excessive computation, especially for high-definition images.
Innovation Solution
An apparatus and method that adaptively reduce noise in image sensors using a kernel for convolution and normalization based on noise characteristics and photography information, with a noise level setting module, kernel generation module, and filter application module, which convert images into a luma-chroma color space and apply kernels controlled by weight functions to filter pixels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If spatial filter with smoothing function is applied to reduce noise, then noise is reduced, but original image properties are damaged
Solution Approach 1:
The patent applies different filtering strengths to different regions of the image based on local noise characteristics. The filtering operation adapts its intensity locally, preserving edges and important features while reducing noise in uniform regions, thus resolving the contradiction between noise reduction and image quality preservation
Solution Approach 2:
The filtering kernel is dynamically adjusted based on local image characteristics and noise levels. Rather than applying a fixed smoothing filter, the system adapts the filter parameters locally, enabling dynamic balance between noise reduction and preservation of original image properties
2Object-affected harmful factors
If low pass filter is applied to suppress high-frequency components, then noise is reduced, but computation complexity increases
Solution Approach 1:
The patent divides the image processing into discrete pixel operations using separable convolution kernels. This segmentation approach allows noise reduction to be performed through simple, repeated local operations rather than complex global filtering, reducing computation complexity while maintaining effectiveness
Solution Approach 2:
The patent uses kernel templates that are generated once and then reused across the entire image. This copying approach eliminates the need for complex computations at each pixel location, as the same kernel patterns are replicated and applied throughout the image, significantly reducing computational burden
Data Source
AI summary
An apparatus and method for reducing noise of an image sensor are provided. The apparatus includes a noise level setting module setting a noise level representing an image frame expressed as a component in a color space; a kernel generation module generating a kernel for filtering pixels that constitute the image frame based on the set noise level; and a filter application module convoluting the pixels using the generated kernel and normalizing the pixels using the convolution result. Since noise generated by the image sensor can be adaptively reduced, the quality of an output image can be improved.