Image Processing Apparatus Memory Bandwidth Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing methods for image data from single plate image sensors with Bayer arrays face challenges in reducing noise and false colors while maintaining image quality, often requiring large memory capacity due to synchronization of frequency bands and potential degradation in correction accuracy.
Innovation Solution
An image processing apparatus that divides image data into multiple frequency bands, synchronizes the second and third bands for efficient storage, and processes the first band without synchronization, using downsampling and high-pass filters to reduce pixel count and memory requirements, while maintaining Bayer array integrity and correcting noise and aberrations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If image data is divided into multiple frequency bands and stored in memory for correction processing, then noise reduction and aberration correction can be performed, but memory capacity requirements increase due to synchronization of all frequency bands
Solution Approach 1:
The image data is segmented into multiple frequency bands (first, second, and third frequency bands) with different resolutions. The first frequency band contains high-resolution data, while the second and third frequency bands contain downsampled low-resolution data. This segmentation allows selective processing and storage, reducing the total memory capacity required while maintaining correction accuracy for regions that need it.
Solution Approach 2:
Different regions of the image data are processed with different quality levels. The first frequency band (high-resolution) is used for regions requiring detailed correction, while the second and third frequency bands (downsampled) are used for regions where lower resolution suffices. This local quality differentiation reduces overall memory requirements while maintaining necessary correction accuracy.
2Manufacturing precision
If all frequency bands are synchronized and stored in memory, then complete image data is available for correction, but processing time and computational load increase
Solution Approach 1:
The second and third frequency bands are downsampled in advance before storage in memory. This preliminary action reduces the data volume that needs to be processed during correction, decreasing computational load and processing time while still providing sufficient information for accurate correction when combined with the first frequency band.
3Measurement precision
If high-resolution image data is stored and processed, then image quality is maintained, but memory bandwidth and storage requirements increase
Solution Approach 1:
Image data is segmented into multiple frequency bands with different resolutions. The first frequency band maintains high resolution for critical regions, while the second and third frequency bands use downsampled low-resolution data. This segmentation reduces total memory bandwidth requirements while preserving image quality where it matters most.
Solution Approach 2:
Different regions are assigned different quality levels based on their importance. High-resolution data in the first frequency band is used for regions requiring detailed representation, while downsampled data in the second and third frequency bands suffices for other regions. This approach maintains necessary image quality while reducing memory bandwidth consumption.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An image processing apparatus can execute appropriate image processing while preventing increase in a capacity of a memory for storing a plurality of pieces of image data of different frequency bands. The image processing apparatus stores first image data having the highest frequency among the plurality of pieces of image data of different frequency bands in a state in which each pixel of the first image data includes a color component signal of any of a plurality of colors, on the memory, and further stores second image data whose frequency band is lower than that of the first image data on the memory, in a state in which a part of or all pixels of the second image data have color component signals of a plurality of colors.