Multi-ISP Raw Image Blending for Seamless Boundary Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
When multiple image signal processors (ISPs) process image signals with different setting values, consistency among generated images decreases, leading to unnatural boundaries and reduced image quality.
Innovation Solution
A method is employed where a raw image signal is split into overlapping regions, processed by different ISPs, and then merged using a blending region determined by comparing register values and pixel differences to ensure seamless integration.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multiple ISPs process image signals with different setting values, then image processing productivity is improved, but image consistency deteriorates
Solution Approach 1:
The image signal is divided into multiple regions, with each region processed by a different ISP according to its specific requirements. The split circuit separates the incoming image signal into first and second regions, directing them to different ISPs while maintaining overall system productivity.
Solution Approach 2:
Different setting values are applied to different regions of the image based on local requirements. Each ISP receives customized setting values appropriate for its assigned region, allowing optimal processing for each area while maintaining global image consistency through the blending region.
2Speed
If multiple ISPs process image signals independently, then processing speed is improved, but image quality deteriorates due to unnatural boundaries
Solution Approach 1:
The blending circuit merges the first and second processed regions by determining a blending region where the two regions overlap. This merging process combines the separately processed regions smoothly, eliminating unnatural boundaries while maintaining the processing speed benefits of multiple ISPs.
Solution Approach 2:
The blending region acts as an intermediary zone between the first and second processed regions. By processing this overlapping blending region through both ISPs and combining the results, the system creates a smooth transition area that eliminates harsh boundaries between regions processed by different ISPs.
3Stability of the object's composition
If overlapping regions are transmitted to multiple ISPs, then image consistency is improved, but device complexity increases
Solution Approach 1:
The split circuit and blending circuit are designed to handle multiple functions: the split circuit both divides the image signal and determines region assignments, while the blending circuit both merges regions and determines blending region boundaries. This multi-functionality reduces the need for separate dedicated components, managing device complexity while maintaining image consistency.
Data Source
AI summary
A method, performed by an image processing device, of processing an image signal is provided. The method includes receiving, by a split circuit, a first raw image signal, which is at least a portion of an image signal captured by an image sensor of a first camera, determining, by the split circuit, a blending region for the first raw image signal, transmitting, from the split circuit, a first partial raw image signal of the first raw image signal corresponding to a first region including the blending region to a first image signal processor (ISP), and transmitting, from the split circuit, a second partial raw image signal of the first raw image signal corresponding to a second region including the blending region to a second ISP. The first partial raw image signal is processed by the first ISP, and the second partial raw image signal is processed by the second ISP.