Hardware-oriented joint denoising and demosaicing method

Abstract

The invention discloses a hardware-oriented joint denoising and demosaicing method, which comprises the following steps of: 1, calculating accurate weights in left, right, upper and lower directions in a window taking red or blue as a center according to the characteristics of Bayer Raw data, and interpolating all missing green information; 2, calculating a high-frequency coefficient required for interpolating a green channel; 3, calculating a global direction weight and a voting direction weight; 4, interpolating the green channel; 5, performing edge-preserving direction wavelet denoising on the green channel; 6, performing interpolation on the color channel by adopting an interpolation matrix designed for the color channel; and 7, performing direction wavelet denoising on the color channel. According to the method, in a low-cost range which can be accepted by a mobile hardware platform, the noise can be removed while the mosaic is removed, so that the image quality of an existing scheme can be improved, and the operation complexity and the power consumption are improved.

Description

technical field [0001] The invention relates to the field of image signal processing, in particular to a hardware-oriented joint denoising and demosaicing method. Background technique [0002] Smartphones have replaced digital cameras and SLR cameras as the primary tool for everyday photography due to their popularity, portability, and powerful processing capabilities. As an important link in the imaging system, the image signal processor determines the quality of the image to a large extent. The image signal processor on the smartphone platform has several major challenges: (1) As a substitute for SLR cameras, smartphones, especially flagship phones, have high requirements for image quality; (2) Due to the limitations of hardware area and power consumption, The requirements for algorithm complexity are also very high on the smart phone platform; (3) The mobile phone platform needs the ability to process high-frame-rate and high-resolution images in real time. In image sig...

AI Technical Summary

Problems solved by technology

[0007] Disadvantage 1: The denoising and demosaic effect of the cascade method is not good

Regardless of whether demosaicing or denoising is first, there are defects, and the known information cannot be fully utilized to obtain the optimal result.

[0008] Disadvantage 2: There are image quality problems in joint denoising and demosaicing based on traditional methods

The existing public joint denoising and demosaicing algorithms have a large gap with the actual use requirements in terms of objective peak signal-to-noise ratio (PSNR) and subjective image quality

Especially in the challenging high-frequency color areas, there are obvious false colors, which cannot meet the actual needs

[0009] Disadvantage 3: The current joint denoising and demosaicing hardware implementation based on traditional methods is too expensive

However, the current algorithm is mainly iterative, and the processing window, computing time and computational complexity cannot meet the requirements.

[0010] Disadvantage 4: The method based on deep learning cannot be implemented on the mobile phone hardware platform

Any of the above conditions limits the implementation of deep learning methods on hardware platforms

Images