Image Processing Apparatus Parallel Edge Extraction Rendering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image processing systems face inefficiencies in rendering speed, particularly when dealing with multiple graphic objects, as they often require replicating data across band regions and cannot perform parallel processing effectively due to dependencies between figures, leading to suboptimal performance.
Innovation Solution
An image processing apparatus that includes a receiving unit for sequential PDL data, an assigning unit for edge extraction processing, a merging unit for merging edge information, a spool unit for data formatting in a tile format, and a generation unit for pixel generation from intermediate data, allowing for independent processing of each figure and efficient distribution across multiple CPU cores.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If figures are divided into band regions for parallel processing, then processing speed can be improved under certain conditions, but figures spanning multiple band boundaries cannot be processed in parallel and require data replication
Solution Approach 1:
The patent segments the processing unit rather than the spatial domain. Each figure is assigned to a specific processing unit (thread) for complete independent processing, eliminating the need for spatial segmentation into bands that causes boundary problems. This segmentation by processing task rather than spatial region allows parallel processing of all figures without data replication.
Solution Approach 2:
The patent introduces an edge information extraction stage as an intermediary between figure data input and rendering. By extracting edge information for each figure independently first, then passing to rendering stage, it enables parallel processing of edge extraction for multiple figures simultaneously without the band boundary constraints that would require data replication.
2Productivity
If one thread is assigned to one band for processing, then parallel processing can be performed, but unbalanced object distribution across bands prevents high-speed processing
Solution Approach 1:
The patent makes the thread-to-figure assignment dynamic rather than static. Processing units can be dynamically assigned to different figures based on current workload and figure complexity, allowing load balancing to adapt to unbalanced object distribution. This dynamic assignment enables efficient utilization of multiple threads regardless of how figures are spatially distributed.
Solution Approach 2:
The patent segments by processing task (assigning each figure to a dedicated thread) rather than by spatial band. This task-based segmentation naturally handles unbalanced distributions because each processing unit works independently on its assigned figure(s) without requiring balanced spatial distribution across fixed bands.
3Productivity
If edge data is generated when figure data is transferred to the renderer, then processing can be performed, but each figure cannot be independently processed in parallel
Solution Approach 1:
The patent extracts the edge information generation step from the rendering pipeline and makes it an independent parallel processing stage. Edge information for each figure is extracted independently by dedicated processing units before being passed to the rendering stage, removing the dependency that prevented parallel processing while maintaining processing efficiency.
Data Source
AI summary
A method for image processing executed by an image processing apparatus includes sequentially receiving PDL data and transferring a figure included in the PDL data to processing in a subsequent stage, assigning edge extraction processing as a unit of processing for extracting edge information for each transferred figure, merging the edge information extracted from each figure, spooling the merged data in a tile format as intermediate data, and reading out the intermediate data in the tile format from spooling process and performing processing on each tile to generate a pixel from the intermediate data.