Imaging Sensor Phase-Difference Pixel Signal Mixing and Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In imaging devices with phase-difference pixels, the use of phase-difference pixels for autofocus leads to insufficient sensitivity, requiring increased gain and interpolation, which increases memory usage and degrades image quality, especially when phase-difference pixels are arranged in both horizontal and vertical directions.
Innovation Solution
An imaging device and method where phase-difference pixels are arranged corresponding to specific color filters, with mixed pixel outputs corrected differently based on their orientation, avoiding the need for widespread pixel arrangements that increase memory usage and maintaining image quality by selectively including or excluding phase-difference pixel signals in mixed pixel outputs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If phase-difference pixels are arranged in a wide area and mixed with imaging pixels to improve sensitivity, then phase-difference detection performance is improved, but memory usage increases and image quality degrades due to the need for wide-area pixel arrangement for gain estimation
Solution Approach 1:
The patent applies local quality by differentiating the treatment of phase-difference pixels based on their position and orientation. Different gain correction values are applied to horizontal phase-difference pixels versus vertical phase-difference pixels, and to pixels at different distances from the image center. This localized approach allows effective phase-difference detection without requiring uniform wide-area pixel arrangements, thereby reducing memory usage while maintaining detection performance.
Solution Approach 2:
The patent changes parameters by dynamically adjusting gain correction values based on pixel position and orientation rather than using a fixed gain for all phase-difference pixels. The gain correction value varies according to the distance from the image center and the orientation (horizontal/vertical) of the phase-difference pixel. This parameter change enables accurate correction with fewer memory resources compared to uniform gain application across the entire sensor area.
2Measurement precision
If gain is increased to correct insufficient sensitivity of phase-difference pixels, then phase-difference detection sensitivity is improved, but image quality degrades due to over-amplification of noise and interpolation errors
Solution Approach 1:
The patent applies parameter changes by using position-dependent and orientation-dependent gain correction values instead of a uniform gain. Pixels closer to the image center receive different gain correction than pixels at the periphery, and horizontal phase-difference pixels receive different correction than vertical ones. This differentiated parameter approach improves sensitivity where needed while avoiding over-amplification in regions where phase-difference detection is less critical, thereby maintaining image quality.
3Measurement precision
If phase-difference pixels are arranged in both horizontal and vertical directions to improve detection capability, then phase-difference detection performance is improved, but the complexity of gain estimation and correction increases significantly
Solution Approach 1:
The patent applies segmentation by dividing phase-difference pixels into distinct groups based on their orientation (horizontal vs. vertical) and position (distance from image center). Each segment is assigned specific gain correction values appropriate to its characteristics. This segmentation simplifies the correction process compared to treating all pixels uniformly, as it allows independent optimization for each orientation and position category, reducing overall calculation complexity.
Solution Approach 2:
The patent applies local quality by implementing direction-dependent gain correction where horizontal phase-difference pixels and vertical phase-difference pixels are corrected differently. Additionally, pixels at different distances from the image center receive different correction values. This localized correction strategy reduces the complexity of gain estimation by treating each local region and orientation independently rather than requiring complex global estimation across the entire sensor array.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach prevents memory increases and image quality degradation while enhancing phase-difference detection performance by optimizing pixel signal mixing and correction, particularly in digital cameras.
Implementation Method 1
an imaging section including a plurality of photodiodes arranged in a matrix configuration... each photodiode generates a photoelectric conversion current corresponding to a light receiving amount
Data Source
Figure 1A
Figure 1B
Figure 2
AI summary
To prevent an increase in memories and a degradation in image quality while improving the phase-difference detection performance. Phase-difference pixels LF/RI are arranged at the positions corresponding to a first color filter (G filter) of an imaging pixel, and phase-difference pixels TP/BT are arranged at the positions corresponding to a second color filter (B filter) different from the first color filter of imaging pixels. When mixing and reading out the outputs of imaging pixels corresponding to the first color filter (Yes, in S51), a plurality of pixel signals including the output of a phase-difference pixel (LF/RI) is mixed and read out as a first mixed pixel output. When mixing and reading out the outputs of the imaging pixels corresponding to the second color filter (No, in S51), a plurality of pixel signals without including the output of a phase-difference pixel (TP/BT) is mixed and read out as a second mixed pixel output. The output value of the first mixed pixel output is corrected (S55), while the output value of the second mixed pixel output is not corrected.