Image Processing Apparatus Pixel Order Attribute Map Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In image processing, especially when combining or applying geometric transformations to images, the correspondence between pixels in the input image and the attribute map is often lost due to preceding processing stages, leading to inefficiencies and increased costs, particularly when handling higher-definition images or performing filter processes on block areas that change size.
Innovation Solution
An image processing apparatus that includes a filter unit, a supply unit, and an image processing unit, where the pixel order for input and output of the filter process differs, and the supply unit provides attribute information based on the position and size of block areas after filtering, ensuring proper correspondence between image pixels and attribute map pixels, thereby preventing the need for overlapping block areas or additional memory buffers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If block areas are enlarged to overlap adjacent blocks to prevent size changes after filter process, then pixel correspondence is maintained, but data transfer efficiency decreases and processing performance lowers
Solution Approach 1:
The patent applies preliminary action by performing geometric transformation on the attribute map in advance, before the filter process, so that the attribute map is transformed in the same way as the input image. This ensures pixel correspondence is maintained without needing to overlap block areas or reduce processing speed, thus resolving the contradiction between reliability and productivity.
2Reliability
If overlapping block areas are used to maintain pixel correspondence, then attribute map alignment is improved, but memory access efficiency decreases
Solution Approach 1:
The attribute map undergoes geometric transformation in advance, before block-based filtering, so that when blocks are processed, the attribute map data is already aligned with the transformed image blocks. This eliminates the need for overlapping blocks and reduces redundant memory accesses, improving memory access efficiency while maintaining alignment.
Solution Approach 2:
The patent segments the attribute map into block-based units that correspond to the image blocks, allowing independent processing of each block with its corresponding attribute data. This segmentation enables efficient memory access patterns without requiring overlapping blocks, thus improving memory access efficiency while maintaining proper alignment.
3Productivity
If filter process is applied to block areas, then processing efficiency is improved, but block area sizes change depending on positions, losing pixel correspondence
Solution Approach 1:
Geometric transformation is applied to the attribute map in advance, before the filter process divides the image into blocks. This preliminary transformation ensures that when blocks are extracted and filtered, the attribute map data corresponds correctly to each block's pixels, maintaining pixel correspondence while allowing efficient block-based processing.
Data Source
AI summary
An image processing apparatus includes: a filter unit configured to execute a filter process on each block area in the input image and output the filter processing result of each pixel in the input image; a supply unit configured to read out attribute information from the attribute map and supply attribute information for each pixel; and an image processing unit configured to perform the predetermined image processing for the filter processing result output by the filter unit based on the attribute information for each pixel supplied by the supply unit, wherein the pixel order in which the pixel value in the input image is received by the filter processing unit and the pixel order in which the filter processing unit outputs the filter processing result are different, and wherein the supply unit supplies the attribute information for each pixel according to the pixel order of the filter processing result.