A raw image data noise calibration and denoising method based on dct transform
By combining 2D and 3D denoising methods and using dct transform for noise calibration and denoising, the problems of high storage cost and inconsistent denoising effect of ISP image denoising methods are solved, achieving efficient image compression and consistent denoising effect.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- INGENIC SEMICON CO LTD
- Filing Date
- 2025-01-15
- Publication Date
- 2026-07-17
AI Technical Summary
In existing technologies, ISP image denoising methods require memory space to store historical frame data, resulting in high storage costs. Furthermore, the noise level varies under different lighting conditions, leading to inconsistent denoising effects that cannot be guaranteed.
Combining 2D and 3D noise reduction methods, the DCT transform is used for noise calibration and reduction. By capturing images of a 24-color card under different lighting conditions, raw data of the sensor at different gain values is obtained. The DCT transform is then performed to fit the mapping relationship between the DC and AC components. The noise is estimated using the fitting formula, and a thresholding method is used for noise reduction. Finally, the DCT transform is used for image compression when compressing historical frames in 3D noise reduction.
It improves image compression rate, ensures consistent noise reduction effect, and reduces storage cost.
Smart Images
Figure CN122415368A_ABST