ML Dot Data Generation for Printer Artifact Suppression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image processing technologies for printers fail to adequately reduce artifacts such as periodicity and false contouring in images printed using binary data, leading to degraded image quality.
Innovation Solution
An image processing device that utilizes a machine learning model to generate dot data for printers by inputting partial image data into the model with varying input parameters, even when the image data remains the same, to produce different dot formation states, thereby suppressing periodicity and false contouring.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional thresholding using neural network is used to generate binary data, then the processing speed is maintained, but artifacts such as periodicity and false contouring appear in printed images
Solution Approach 1:
The patent applies dynamics by making the dot formation state variable rather than fixed. The machine learning model generates different dot patterns dynamically based on the input parameter, even for the same image data. This dynamic adjustment suppresses periodicity and false contouring artifacts while maintaining processing efficiency through automated parameter variation.
Solution Approach 2:
The patent changes the parameter of dot formation state by introducing an input parameter to the machine learning model. By varying this parameter, the system generates different dot patterns from the same input image data, which effectively reduces artifacts like periodicity and false contouring while maintaining binary data processing capabilities.
2Stability of the object's composition
If the same dot formation state is used for identical image data, then processing consistency is maintained, but periodicity artifacts appear in the printed image
Solution Approach 1:
The system maintains stability in processing consistency while introducing dynamic variation in dot formation states. The machine learning model consistently processes identical image data through the same workflow, but varies the output dot patterns by adjusting the input parameter, thereby preventing periodicity artifacts without compromising processing reliability.
Solution Approach 2:
By changing the input parameter that controls dot formation state, the system generates varied dot patterns from identical image data. This parameter variation breaks the periodicity that would otherwise occur with repeated identical patterns, while the consistent application of the machine learning model maintains processing stability.
3Device complexity
If binary data with fixed dot patterns is used for printing, then device complexity is reduced, but false contouring artifacts degrade image quality
Solution Approach 1:
The patent introduces parameter variation in dot formation states without complicating the device structure. By adjusting the input parameter to the machine learning model, the system generates different dot patterns that suppress false contouring artifacts, maintaining binary data simplicity while improving image quality through intelligent parameter control.
Data Source
AI summary
In an image processing device, a controller performs: acquiring target image data representing a target image; inputting first and second datasets into a machine learning model and causing the machine learning model to output first and second partial dot data; and generating dot data specifying a dot formation state for each of a plurality of pixels in a print image corresponding to the target image using the first and second partial dot data. The first dataset includes first partial image data and a first value for an input parameter. The second dataset includes second partial image data and a second value for the input parameter. The machine learning model outputs the second partial dot data different from the first partial dot data according to the second value being different from the first value even when the second partial image data is identical to the first partial image data.