Blob Encoding for Non-Contiguous Printing Defect Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for detecting printing errors in digital images fail to recognize spatially extended and non-pixel-wise coherent image features, leading to inaccurate classification of printing defects, as they rely solely on direct neighborhood relationships and cannot combine non-contiguous foreground pixels effectively.
Innovation Solution
A method that adapts the surroundings of individual pixels to combine spatially extended image features by considering marked surrounding pixels with different markings as equivalent, using a predefined sequence to search for marked pixels in sub-environments, and assigning a new marking if none is found, allowing for the identification of coherent image features across varying distances.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional blob encoding methods using direct neighborhood relationships are used, then the detection process is simple and fast, but spatially extended and non-pixel-wise coherent image features cannot be recognized
Solution Approach 1:
The patent segments the image processing into multiple passes, with each pass examining pixels within a specific distance range from already-marked pixels. This segmentation allows the system to handle non-contiguous pixels by processing them in staged increments rather than requiring all pixels to be directly connected, thereby improving defect detection accuracy while maintaining manageable computational complexity.
2Reliability
If the resolution of the digital image is reduced to enlarge defective areas, then non-contiguous blobs become visible, but measurement precision deteriorates
Solution Approach 1:
Instead of reducing spatial resolution to make non-contiguous defects visible, the patent introduces a temporal dimension by processing the image in multiple passes. Each pass extends the examination distance from marked pixels, effectively adding a dimension of processing depth without compromising the original spatial resolution and measurement precision.
3Reliability
If multiple small separate blobs are detected individually, then each blob can be analyzed, but clusters of non-contiguous blobs representing a single printing error cannot be recognized
Solution Approach 1:
The patent implements continuity of useful action by performing multiple sequential passes over the same image data. Each pass continues the marking process from where previous passes left off, examining pixels at increasing distance ranges from already-marked pixels. This continuous multi-pass approach efficiently clusters non-contiguous blobs that belong to the same printing error without requiring repeated image acquisition or complex post-processing.
Data Source
Figure 1a~3
Figure 3a~3c
Figure 4
AI summary
The invention relates to a method for identifying spatially extended and image features (F) that are not interrelated in terms of pixels in two-dimensional digital images (I) to be examined having pixels (P) arranged in a grid-type manner in the form of a rectangular grid, wherein two coordinate values (x, y) are attributed in each case to each pixel (P) of the two-dimensional digital image (I) according to the position thereof in the rectangular grid of the digital image (I), and the respective coordinate values (x, y) of adjacent pixels (P) differ in each case by one pixel dimension, wherein the individual pixels are passed through in the sequence of a regular scan, wherein for each pixel (P) that is run through in such a manner, the following method steps a) to d) are carried out, namely that the color and/or brightness values attributed to the respective pixels are examined with respect to the presence of a predetermined criterion, in particular with respect to exceeding or dropping below a predetermined threshold value (T), when this criterion is present an environment (U) of a predetermined size and shape surrounding the respective pixel (P) is determined, wherein pixels are only added to the environment (U) as environment pixels (Pu) when said environment pixels (Pu) have already been examined with respect to marked pixels (P) in the environment thereof, at least one environment pixel (Pu) is attributed to the environment (U) of the respective pixel (P), wherein at least one of the coordinate values (x, y) of the environment pixel (Pu) differs from the respective coordinate value (x, y) of the pixel (P) by at least two pixel dimensions and within the environment (U) it is searched for already marked environment pixels (Pu), and when a minimum number of marked environment pixels (Pu) is marked with the same marking (Ml…Mn) within the environment (U), in particular of at least one marked environment pixel (Pu), the same marking (Ml…Mn) is attributed to the respective pixel (P), otherwise a new, and not attributed marking (Ml…Mn) is attributed to the respective pixel (P), and subsequently after the passage of the pixel (P) the amounts of pixel (P) to which in each case the same marking (Ml…Mn) has been attributed, are considered as image features (F).