A Raw Image Denoising Method Based on Sparse Representation

Abstract

The invention discloses a RAW image denoising method based on sparse representation, which includes the following steps: Step 1: Decompose the RAW image to be denoised into overlapping RAW rectangular image blocks of fixed size; Step 2: Divide the RAW image blocks obtained in Step 1 into RAW rectangular image blocks are rearranged into G1RBG2 color layers according to different color channels; Step 3: Stretch the G1RBG2 color layer obtained in Step 2 in the order of G1, R, B, and G2 to obtain an image block vector in the form of a column vector , splice each image block vector from left to right into a G1RBG2 matrix, and each column of the G1RBG2 matrix is ​​a training sample; Step 4: Randomly select some training samples from the G1RBG2 matrix in step 3 as the training sample set T, using K -SVD method learns the training sample set T to obtain the dictionary D; Step 5: Use the improved OMP algorithm to reconstruct the dictionary D obtained in Step 4 to form a noiseless RAW image; this method solves the problem Computationally complex issues in RAW image denoising.

Description

【Technical field】 [0001] The invention belongs to the technical field of image processing, in particular to a method for denoising RAW images based on sparse representation. 【Background technique】 [0002] Compared with large satellites, microsatellites have the characteristics of light weight, low cost, and short design and manufacturing cycles. As an important task of space applications, Earth observation has a wide range of applications, such as vegetation monitoring and disaster monitoring. In this field, microsatellites have been successfully applied. Image processing systems are a key component of Earth observation missions. [0003] However, due to the complex space environment, the images captured by the camera are severely affected by noise. For example, weak light intensity will bring a lot of noise. The camera amplifies the voltage of the photosensitive element by increasing the sensitivity of the input signal (increasing ISO), which will cause noise enhancemen...

AI Technical Summary

Problems solved by technology

[0006] The purpose of the present invention is to provide a method for denoising RAW images based on sparse representation, so as to solve the problem that the noise parameters cannot be effectively estimated and the calculation is complicated in the process of processing undenoised RAW images.

Images