Painterly Image Conversion via Vector Coherency Deformation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing techniques for achieving a painterly effect, such as those described in JP H8-44867A and Non-patent Document 1, require repetitive processes to obtain a desirable output, leading to long processing times, while techniques like JP 2011-160306A and JP 2011-166217A struggle with brush touch control, resulting in inefficient image conversion.
Innovation Solution
An image processing apparatus and method that includes a vector detection unit, a vector-coherency calculation unit, a deformation-characteristic computation unit, and a painterly conversion unit, which detect flow vectors, calculate vector coherency, compute deformation characteristics, and use these to deform filter shapes for each pixel, allowing for a single-process painterly conversion of images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If repetitive brush drawing processes are used to obtain desirable painterly output, then painting style control is improved, but processing time increases significantly
Solution Approach 1:
The patent pre-calculates and stores deformation characteristics for various brush types, sizes, and strokes in lookup tables before actual image processing. When converting an image, the system simply retrieves pre-computed deformation characteristics based on selected brush parameters, eliminating the need for repetitive iterative drawing processes and achieving both precise brush control and fast processing
Solution Approach 2:
The patent transforms the complex repetitive drawing process into a parameter-based retrieval system. By changing the approach from iterative application to direct parameter selection (brush type, size, stroke) with pre-computed deformation characteristics, the system achieves efficient single-pass image conversion while maintaining artistic control
2Productivity
If brush size is changed dynamically based on distance or area, then processing speed is improved, but brush touch control suitable for painting becomes difficult
Solution Approach 1:
The patent applies different deformation characteristics to different regions of the image based on local features detected by the feature detection unit. Each region receives appropriately tailored brush deformation parameters (size, shape, stroke) that match the local image characteristics, achieving both efficient processing and artistically appropriate brush touches
Solution Approach 2:
The system uses feature detection to analyze image characteristics and provides feedback for selecting appropriate deformation characteristics from pre-computed tables. This feedback mechanism ensures that brush parameters are automatically adapted to match the content and style requirements of different image regions without manual intervention
Data Source
AI summary
There is provided an image processing apparatus including a vector detection unit which detects flow vectors of pixels in an inputted image, a vector-coherency calculation unit which calculates vector coherency based on the flow vectors detected by the vector detection unit, a deformation-characteristic computation unit which computes a deformation characteristic by using the vector coherency calculated by the vector-coherency calculation unit, the deformation characteristic being used for deforming a tap shape of a filter used for each of the pixels, and a painterly conversion unit which converts the inputted image based on the deformation characteristic computed by the deformation-characteristic computation unit.