Image Processing PSNR Diagnostic via Noise-Added Reference
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In automatic driving systems, the low peak signal to noise ratio (PSNR) between original images and their decoded images, often below a preset threshold, indicates poor decoded image quality, which can hinder normal operation, and current reasons for this issue are speculative without theoretical basis.
Innovation Solution
An image processing method that involves obtaining a PSNR between an original image and its decoded image, and a noise-added image and its decoded image, to determine the credible reason for low PSNR, by encoding and decoding both images and analyzing the differences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If image compression and encoding is performed to reduce data transmission, then transmission efficiency is improved, but image quality deteriorates (lower PSNR)
Solution Approach 1:
The patent changes the parameter being measured from absolute image quality (PSNR between original and decoded image) to relative quality improvement (PSNR between noise-added image and decoded image). This allows the system to accept lower absolute quality in exchange for higher transmission efficiency, while still ensuring the decoded image is significantly better than the noisy input.
2Manufacturing precision
If noise reduction processing is applied to improve image quality, then image quality is improved, but processing complexity increases
Solution Approach 1:
The patent creates a synthetic noise-added copy of the original image by adding artificial noise. This copy serves as a reference to measure whether the decoding process actually improves quality, without requiring complex noise reduction algorithms. The method uses simple noise addition rather than complex noise reduction to achieve its quality assessment goal.
3Manufacturing precision
If PSNR threshold checking is performed to ensure image quality, then image quality control is improved, but diagnostic capability deteriorates (cannot identify cause of low quality)
Solution Approach 1:
The patent segments the quality assessment into two distinct measurements: one for the original image pathway and one for the noise-added image pathway. By comparing results from these separate segments, the system can diagnose whether low quality stems from the encoding process itself or from inadequate noise handling, thereby improving diagnostic capability while maintaining quality control.
Data Source
AI summary
The present application discloses an image processing method, apparatus, device and storage media. A specific implementation solution is: obtaining an original image and a noise-added image, where the noise-added image is an image of the original image after a noise is added, and a number of pixels with a noise in the original image is less than a preset number; encoding and decoding the original image and the noise-added image respectively, to obtain a first decoded image corresponding to the original image and a second decoded image corresponding to the noise-added image; obtaining a first PSNR between the first decoded image and the original image according to the original image and the first decoded image; obtaining a second PSNR between the second decoded image and the noise-added image according to the noise-added image and the second decoded image; and outputting the first PSNR and the second PSNR.