Rain Removal Image Processing Using Color Space Signal Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image processing systems require a large capacity image memory to calculate the average value of consecutive frames to remove the influence of rain or similar elements, leading to memory inefficiencies.
Innovation Solution
An image processing apparatus that selects two image signals with a shorter distance in the color space from input, immediately preceding, and already output signals, and generates an output signal by mixing them based on predetermined ratios, reducing the need for extensive memory storage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If images of at least three frames are held in image memory to calculate average value, then the influence of rain or moving objects is removed, but image memory capacity increases
Solution Approach 1:
The patent extracts only the necessary components for rain removal by using a difference image generation unit that computes the difference between current and reference frames, rather than storing complete frame images. This extraction approach retains only the essential difference data needed for identifying raindrops, significantly reducing memory requirements while maintaining rain removal effectiveness
Solution Approach 2:
The patent segments the image processing task into distinct functional units: difference image generation, raindrop detection based on saturation, and background reconstruction. By segmenting the process, the system processes only relevant data portions rather than entire frames, reducing the memory burden while preserving the ability to remove rain influence
2Reliability
If images of consecutive frames are held to calculate average value, then background image is generated, but processing time increases
Solution Approach 1:
The patent applies preliminary action by pre-selecting reference frames and pre-generating difference images before final rain removal processing. The reference frame selection unit identifies suitable reference frames in advance, and the difference image generation unit prepares difference data beforehand, reducing the computational burden during real-time processing and decreasing overall processing time
Solution Approach 2:
The patent uses partial action by focusing processing only on regions where raindrops are detected rather than processing entire frames. The raindrop detection unit identifies specific areas requiring processing, and subsequent operations are applied only to those regions, reducing total processing time while maintaining background image accuracy
Data Source
AI summary
An image memory is reduced in an image processing apparatus that generates a background image from which the influence of rain or the like is removed.In the image processing apparatus, a selection unit selects two image signals having a shorter distance in a color space from among an input image signal input during a focused processing cycle among processing cycles executed for respective time-series frames each constituted by the input image signal, an immediately preceding output image signal which is an output image signal generated during a processing cycle immediately preceding the focused processing cycle, and an already output image signal which is an image signal based on the output image signal generated during a processing cycle preceding the focused processing cycle by two cycles. An output image signal generation unit generates the output image signal by mixing the two selected image signals on the basis of a predetermined mixing ratio.