Bayer Image Encoding Segmentation for Noise Suppression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image encoding techniques for Bayer arrangements, such as JPEG, result in inefficient compression due to folding noise and low compression ratios, particularly when handling green component pixels, as they do not effectively separate and encode the high and low-frequency components of the G component.
Innovation Solution
An image encoding apparatus that generates low-frequency (GL) and high-frequency (GH) components from G0 and G1 data, and performs luminance/color difference transformation on R, B, and GL data to create Y, U, and V planes, allowing for efficient encoding of these components using wavelet transformation and entropy encoding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If image data of Bayer arrangement is encoded without demosaicing (as in literature 1), then recording efficiency is improved, but folding noise is applied and compression ratio is reduced
Solution Approach 1:
The patent segments the green component pixels (G0 and G1) into separate low-frequency (GL) and high-frequency (GH) components through wavelet transformation. This segmentation allows the high-frequency component to be processed separately, preventing folding noise from contaminating the low-frequency component while maintaining efficient compression. The GL component is then used for luminance/color difference transformation, and the GH component is encoded separately to preserve compression efficiency.
2Productivity
If G0 and G1 components are separated into different components (as in literature 1), then encoding without demosaicing is achieved, but folding noise is applied when performing wavelet transformation
Solution Approach 1:
The patent applies wavelet transformation to the combined G0 and G1 components before separating them into GL and GH components. This preliminary action of transforming the combined data first prevents folding noise from being introduced during subsequent separation operations. The low-frequency component (GL) is then extracted and used for luminance/color difference transformation, while the high-frequency component (GH) is encoded separately.
3Ease of operation
If JPEG encoding is applied to image data after demosaicing, then color space transformation to YUV is performed, but data amount becomes three-times larger than Bayer arrangement data
Solution Approach 1:
The patent extracts only the necessary green low-frequency component (GL) from the G0 and G1 pixels to perform luminance/color difference transformation. Instead of fully demosaicing all pixels to create complete RGB images, the method extracts only the essential GL component needed for YUV transformation. This extraction approach maintains the compact Bayer arrangement data structure while enabling efficient YUV conversion, keeping the data amount close to the original Bayer data rather than tripling it.
Data Source
AI summary
This invention enables compression-coding of image data of a Bayer arrangement more efficiently. For this purpose, an encoding apparatus includes a generation unit which generates, from G0 and G1 component data of the image data of the Bayer arrangement, a GL plane formed from low-frequency component data of a G component and a GH plane formed from high-frequency component data of the G component, a luminance/color difference transforming unit which generates, from R and B component data of the image data of the Bayer arrangement and the GL plane, a luminance plane, a first color difference plane, and a second color difference plane, and an encoding unit which encodes the luminance plane, the first color difference plane, the second color difference plane, and the GH plane.