Image Quantization for High-Frequency Pattern Density Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image quantization methods fail to maintain density as an area in high-frequency patterns, such as hatching patterns, due to low error propagation in the error diffusion process, leading to abrupt density changes and reversal between light and dark parts.
Innovation Solution
A hybrid quantization process combining dither matrix noise and accumulated error diffusion, using adjustable noise and error use rates to propagate unused errors and adjust the influence of both processes, ensuring density maintenance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If the error diffusion process is used with low error use rate, then texture and dot delay are suppressed, but density as an area cannot be maintained in high-frequency patterns
Solution Approach 1:
The patent combines dither process and error diffusion process into a hybrid quantization method. The dither matrix noise and accumulated error are both applied to each pixel with adjustable weights, merging the advantages of both processes: dither provides good texture suppression while error diffusion maintains density, and together they achieve both goals simultaneously
Solution Approach 2:
The patent introduces error use rate and noise use rate as adjustable parameters that control the degree of influence of accumulated error and dither matrix noise respectively. By dynamically adjusting these parameters based on image characteristics and processing stage, the system optimizes the balance between texture suppression and density maintenance for different types of patterns
2Manufacturing precision
If the error diffusion process is used with high error use rate, then density as an area is maintained, but texture and dot delay occur
Solution Approach 1:
The patent combines dither process and error diffusion process into a hybrid quantization method. The dither matrix noise and accumulated error are both applied to each pixel with adjustable weights, merging the advantages of both processes: dither provides good texture suppression while error diffusion maintains density, and together they achieve both goals simultaneously
Solution Approach 2:
The patent introduces error use rate and noise use rate as adjustable parameters that control the degree of influence of accumulated error and dither matrix noise respectively. By dynamically adjusting these parameters based on image characteristics and processing stage, the system optimizes the balance between texture suppression and density maintenance for different types of patterns
3Object-affected harmful factors
If dither process is used, then texture is suppressed, but density as an area cannot be maintained in high-frequency patterns
Solution Approach 1:
The patent combines dither process and error diffusion process into a hybrid quantization method. The dither matrix noise and accumulated error are both applied to each pixel with adjustable weights, merging the advantages of both processes: dither provides good texture suppression while error diffusion maintains density, and together they achieve both goals simultaneously
Data Source
AI summary
An object of the present disclosure is to maintain density as an area as compared to a conventional technology even in a case where the quantization target is an image of a high-frequency pattern, such as a hatching pattern, with a halftone value to be subjected to an error diffusion process. An embodiment of the present invention is a program for causing a computer to function as an image processing apparatus for quantizing a pixel value of each pixel in an input image, the computer being caused to function as: an adding unit configured to add at least part of a difference between an accumulated pixel value and an applied error value for a pixel of interest to a quantization error value obtained by the quantization; and an error distributing unit configured to diffuse an added error obtained by the adding unit to peripheral pixels around the pixel of interest.