Encoder-Decoder Network for Color Consistency in Low-Light Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Image denoising and enhancement techniques struggle with color consistency issues, particularly under low-light or underwater conditions, leading to inconsistent colors in images due to low signal-to-noise ratio, low contrast, and narrow dynamic range, resulting in artifacts like halos and color inconsistencies.
Innovation Solution
An encoder-decoder network with specific convolutional blocks and receptive fields processes images to encode color consistency relationships into global and local information, ensuring consistent color restoration across image portions by utilizing dilated and vanilla convolutional layers with varying receptive fields.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If image denoising and enhancement techniques are applied under low-light or underwater conditions, then image quality is improved, but color consistency deteriorates due to low signal-to-noise ratio and narrow dynamic range
Solution Approach 1:
The image processing is divided into separate encoding and decoding stages. The encoder segment extracts features and encodes color consistency relationships, while the decoder segment restores colors based on encoded information. This segmentation allows independent optimization of each stage to address both image quality and color consistency.
Solution Approach 2:
An encoder-decoder network acts as an intermediary between the input image and output image. The encoder transforms the input image into encoded representations that preserve color consistency relationships, and the decoder transforms these representations back to image form while maintaining color consistency. This intermediary structure enables controlled color restoration.
2Device complexity
If conventional image processing is used, then processing simplicity is maintained, but color inconsistency artifacts like halos occur
Solution Approach 1:
The encoder-decoder network uses feedback mechanisms where the encoder analyzes the input image to identify color consistency relationships, encodes this information, and the decoder uses this feedback to guide the color restoration process. This feedback loop ensures that color inconsistencies are detected and corrected systematically.
Solution Approach 2:
The network changes parameters such as receptive field sizes (using both 3×3 and 7×7 convolutional layers) and encoding dimensions to optimally capture color consistency relationships. By adjusting these parameters, the system can effectively restore colors while minimizing artifacts like halos.
Data Source
AI summary
A method includes receiving and processing a first image by an encoder-decoder network. The first image includes a first portion and a second portion located at different locations. The encoder-decoder network includes an encoder and a decoder. The encoder is configured to output at least one feature map including global information and local information such that whether a color consistency relationship between the first portion and the second portion of the first image exists is encoded into the global information and the local information. The decoder is configured to output a second image generated from the at least one feature map, wherein a first portion of the second image corresponding to the first portion of the first image and a second portion of the second image corresponding to the second portion of the first image are restored considering whether the color consistency relationship exists.