Motion Vector Reliability Estimation for Wide Dynamic Range Image Synthesis
Solution Overview
Problem
Existing image synthesis techniques for creating wide dynamic range images from continuous shooting under different exposure conditions face challenges in accurately aligning and synthesizing images, particularly when exposure differences are significant, leading to potential blurs and noise due to handshaking or object movement.
Innovation Solution
An image processing apparatus and method that computes motion vectors between multiple images, determines their reliability, and substitutes low-reliability vectors with estimated vectors based on shooting intervals and corresponding high-reliability vectors, ensuring precise position alignment and synthesis, even with large exposure differences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If template matching method is used for position alignment, then computation amount is reduced, but matching precision deteriorates when images have different brightness levels
Solution Approach 1:
The patent applies preliminary action by performing brightness correction on images before template matching. The system calculates a brightness correction value based on exposure time differences and applies it to normalize brightness levels across images. This preliminary brightness normalization enables the use of computationally efficient SAD/SSD methods while maintaining matching precision that would otherwise require the more computationally intensive NCC method.
2Adaptability or versatility
If images with large exposure differences are synthesized, then wide dynamic range is achieved, but positional alignment accuracy deteriorates due to handshaking or object movement
Solution Approach 1:
The patent applies parameter changes by introducing a reliability determination mechanism that evaluates motion vectors based on multiple factors including brightness correction values, exposure time differences, and matching index values. The system dynamically adjusts the reliability assessment parameters to account for handshaking and object movement, enabling accurate positional alignment even when synthesizing images with large exposure differences for wide dynamic range.
3Measurement precision
If motion vector computation is performed between images with large exposure differences, then position alignment is achieved, but reliability of motion vectors deteriorates
Solution Approach 1:
The patent applies feedback by implementing a reliability determination process that uses feedback loops to assess motion vector quality. The system calculates reliability based on feedback from brightness correction values, exposure time differences, and matching index values. When reliability is determined to be low, the system performs reliability improvement processing by selecting alternative motion vectors or adjusting computation parameters, ensuring high-quality position alignment.
Data Source
AI summary
An image processing apparatus obtains a motion vector between at least three images including first, second, and third images obtained through continuous shooting under different exposure conditions and computes a first motion vector between the first and second images and a second motion vector between the first and third images in a plurality of positions on the first image. A reliability level of the motion vector is determined, and an estimated motion vector between the first and second images is computed for the first motion vector having low reliability based on the second motion vector computed in the corresponding position and the shooting interval between the images. In a case where the second motion vector computed in a position corresponding to that of the first motion vector having low reliability has high reliability, the first motion vector having low reliability is substituted with the estimated motion vector.