Defective Pixel Correction in Raw Image Frames
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Raw images from digital cameras often contain defective image frame pixel values due to defective image sensor pixels, which are difficult to detect and correct, especially since de-noising algorithms can smear these defects, making them harder to identify and correct in de-noised images.
Innovation Solution
A method and apparatus that access a digital raw image frame, de-noise it without modifying defective pixel values, detect and correct defective pixel locations, and optionally perform additional de-noising on the corrected frame using a different procedure, employing bilateral filtering and threshold-based detection for defective pixels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If de-noising algorithms are applied to raw images, then image quality is improved, but defective pixel values are smeared with neighboring pixels making detection difficult
Solution Approach 1:
The patent applies de-noising to a first set of image frames before detecting defective pixels, but does so in a way that preserves the ability to detect defects. The de-noised frames are used to generate a defective pixel map that is then applied to correct the original raw frames, effectively performing the defect correction before final noise reduction is needed.
Solution Approach 2:
The patent divides the image processing into separate stages: first de-noising a subset of frames to create a reference, then detecting defective pixels based on that reference, and finally correcting defective pixels in the original frames. This segmentation allows each stage to optimize for its specific purpose without interfering with the others.
2Reliability
If de-noising is performed on raw images, then noise is reduced, but defective pixel values become harder to distinguish from corrected pixels
Solution Approach 1:
The patent performs defective pixel detection and correction on the original raw frames before final image output, using a separate de-noised reference only for detection purposes. This ensures that the actual correction process works on the original high-fidelity data rather than on already-smoothed data where defects are less distinguishable.
Solution Approach 2:
The patent uses a de-noised reference image as an intermediary to identify defective pixels, but does not apply this reference directly to correct the defective pixels. Instead, it uses the reference only to generate a defective pixel map that guides correction of the original frames, preserving the original pixel values for accurate correction.
Data Source
AI summary
In one example, at least a portion of a digital raw image frame captured by a digital image sensor is accessed. The accessed at least a portion of the digital raw image frame is de-noised without substantially modifying defective pixel values when present. In response to determining that at least one image frame pixel in the de-noised at least a portion of the digital raw image frame has a defective pixel value: the locations of each of the at least one image frame pixel having a defective pixel value are detected, and each defective pixel value in each detected location is corrected in the de-noised at least a portion of the digital raw image frame or the originally accessed at least a portion of the digital raw image frame.