Image Processing with Enlargement for Integer-Pixel Motion Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing techniques struggle to accurately detect moving object regions when images are composited with shifts in shooting positions that are non-integer multiples of the pixel pitch, leading to unnatural composite images due to imperfect difference calculation.
Innovation Solution
The method involves obtaining and enlarging two shot images such that the pixel shift between them becomes an integer multiple of the pixel pitch, allowing accurate alignment and difference calculation between the enlarged images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If images are composited with shooting positions shifted by a non-integer multiple of pixel pitch, then high-resolution composite images can be generated, but accurate detection of moving object regions becomes difficult due to residual pixel shifts
Solution Approach 1:
The patent applies preliminary action by performing enlargement processing on images before compositing them. By enlarging images shot at positions shifted by a non-integer multiple of pixel pitch, the system converts the non-integer pixel shift into an integer pixel shift in the enlarged images, enabling accurate alignment and subsequent moving object detection. This preliminary enlargement step resolves the alignment problem before the compositing operation.
Solution Approach 2:
The patent changes the scale parameter of images through enlargement processing. By changing the magnification parameter, the system transforms the pixel shift amount from a non-integer multiple of pixel pitch to an integer multiple, making the images alignable for accurate difference calculation and moving object detection.
2Ease of manufacture
If pixel positions are moved to align images shot at different positions, then compositing can be performed, but residual shifts smaller than pixel pitch remain when shift is a non-integer multiple, reducing difference calculation accuracy
Solution Approach 1:
The patent performs enlargement processing as a preliminary step before compositing. This preliminary action scales the images so that the pixel shift becomes an integer multiple, allowing perfect alignment when pixel positions are moved, and eliminating residual sub-pixel shifts that would otherwise degrade difference calculation accuracy.
Solution Approach 2:
The patent introduces a scaling dimension by enlarging images. This dimensional transformation converts the problem from one where pixel shifts are non-integer multiples to one where they become integer multiples, enabling exact alignment in the discrete pixel domain after enlargement.
Data Source
AI summary
There is provided an image processing apparatus comprising. A first obtaining unit obtains two shot images shot at two respective shooting positions of an image sensor shifted in a first direction by a non-integer multiple of a pixel pitch of the image sensor. A first generating unit generates two enlarged images by enlarging the two shot images such that a pixel shift amount between the two enlarged images, which is derived from a shift between the two shooting positions, becomes an integer number of pixels. A second obtaining unit obtains a difference between the two enlarged images that have been aligned on a basis of the pixel shift amount between the two enlarged images.