Parallel Image Readout for Object Detection and Resolution Conversion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing methods for target object detection, such as face detection, incur high memory access counts due to repetitive read and write operations when generating pyramid images, leading to increased processing time and memory access.
Innovation Solution
An image processing apparatus that reads image data in parallel to both a detection unit and a conversion unit, allowing simultaneous resize and detection processing, thereby reducing memory access counts by outputting readout image data to both units and writing converted data back to storage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If image data is read sequentially for resize processing and detection processing, then processing can be completed in order, but memory access count increases and processing time increases
Solution Approach 1:
The patent applies preliminary action by reading image data from storage into a buffer memory before processing. This allows the data to be readily available for parallel resize and detection operations, eliminating repeated read operations and reducing overall processing time while maintaining correct processing order
Solution Approach 2:
The patent introduces a time dimension to the processing by executing resize processing and detection processing in parallel rather than sequentially. This dimensional change from sequential to parallel execution reduces total processing time while the buffer memory ensures data availability for both operations
2Adaptability or versatility
If pyramid images are generated by sequential resize operations, then images at multiple resolutions are produced, but repetitive read and write operations increase memory access count
Solution Approach 1:
The patent merges the read operations for resize processing and detection processing into a single operation by reading image data once into buffer memory, then making it available to both processing paths. This eliminates repetitive read operations and reduces memory access count while still producing pyramid images at multiple resolutions
Solution Approach 2:
The patent uses buffer memory as an intermediate copy of the image data. Instead of repeatedly reading from and writing to storage, the data is copied once into buffer memory and then processed multiple times from there, significantly reducing storage access operations
Data Source
AI summary
An image processing apparatus comprising a storage unit configured to store image data; a readout unit configured to read out the image data stored in the storage unit; a detection unit configured to detect a target object from the image data read out by the readout unit; a conversion unit configured to convert a resolution of the image data read out by the readout unit; and a write unit configured to write the image data having the resolution converted by the conversion unit in the storage unit, wherein the readout unit outputs the readout image data in parallel to the detection unit and the conversion unit.