Random Spray Retinex Video Denoising via Structured Pixel Sets
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video image enhancement technologies face challenges in denoising and enhancing images, particularly in low-light conditions and with directional halos, and are computationally complex, making them ineffective for real-time video processing.
Innovation Solution
The method employs random spray retinex with structured spray pixel sets and tuned parameters, using low pass filters and blur channels to calculate brightness variations, which are then used to fuse channels and produce denoised and enhanced video images, addressing directional halos and complexity issues.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional retinex algorithm is used for image enhancement, then color constancy and dynamic range compression are improved, but computational complexity increases and denoising effectiveness decreases
Solution Approach 1:
The patent segments the image enhancement process into multiple independent modules: spray pixel set generation, brightness calculation, denoising via low pass filters, and channel fusion. Each module processes specific aspects of the image separately, reducing the computational burden of the overall retinex algorithm while maintaining enhancement effectiveness.
Solution Approach 2:
The patent introduces tunable parameters including spray radius, radius density function, quantity of spray pixel sets, and quantity of pixels. By optimizing these parameters, the algorithm achieves effective denoising and enhancement with reduced computational complexity, resolving the contradiction between reliability and device complexity.
2Reliability
If traditional retinex algorithm is used for image enhancement, then color constancy is improved, but dependency on geometric paths and sample noises increases
Solution Approach 1:
The patent extracts the harmful dependency on geometric paths and sample noises by replacing them with a spray pixel set approach that uses radial distance from the center pixel. This extraction eliminates the problematic dependencies while preserving the essential color constancy function through the brightness calculation and channel fusion mechanisms.
Solution Approach 2:
The patent introduces an intermediary spray pixel set mechanism that mediates between the center pixel and surrounding pixels. This intermediary structure replaces the direct dependency on geometric paths and sample noises with a controlled sampling process that achieves color constancy without the harmful dependencies.
3Productivity
If video image enhancement is performed in real-time, then processing speed is improved, but image quality and denoising effectiveness deteriorate
Solution Approach 1:
The patent applies partial action by processing only the necessary portions of the image through the spray pixel set mechanism rather than the entire image at full resolution. This selective processing maintains real-time speed while achieving sufficient image quality and denoising effectiveness through the optimized spray radius and pixel quantity parameters.
Data Source
AI summary
The invention relates to image processing technology field, and discloses a video image denoising and enhancing method based on random spray retinex, including: structuring spray pixel sets, and tuning parameters related to the random spray retinex based on the spray pixel sets, wherein the parameters include quantity of the spray pixel sets and quantity of pixels; processing video images with random spray retinex based on tuned parameters; denoising the video images processed by the random spray retinex via low pass filters and blur channels to get a brightness variation calculating formula; obtaining a brightness calculating formula of output images, combined with the brightness variation calculating formula, and calculating brightness variations of three channels via the brightness calculating formula to get local brightness estimating vectors; and fusing the three channels based on the local brightness estimating vectors to get denoised and enhanced video images.