Image Rotation Memory Matrix Organization

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Graphics display systems face inefficiencies in image rotation due to the need for RAS and precharge cycles when fetching data from DRAM or SDRAM memory, as image data must be accessed from multiple rows, reducing the efficiency of data fetches.

Innovation Solution

A method and apparatus for efficient image rotation by mapping image data into a memory matrix with specific vector portions across memory rows and columns, allowing for efficient fetching of data regardless of rotation, using a memory storage scheme that groups pixels to minimize page changes during memory access.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If image data is stored in raster order in DRAM/SDRAM memory, then data fetching efficiency for non-rotated images is improved, but the number of RAS and precharge cycles increases when rotating images

Engineering Contradiction:
Improvedata fetching efficiencyVSAvoidRAS and precharge cycles
Core Design Contradiction:
ProductivityVSLoss of time

Solution Approach 1:

The patent applies preliminary action by pre-organizing image data in memory during the storage phase. The memory is configured to store pixels in an interleaved pattern where pixels from different rows are alternately placed (e.g., row 0, row 2, row 4 in one sequence and row 1, row 3, row 5 in another). This preliminary arrangement ensures that when rotation is needed, the data can be fetched efficiently without requiring multiple RAS and precharge cycles, thus resolving the contradiction between non-rotated image efficiency and rotated image performance.

Inventive Principle:
Principle #10Preliminary action

2Adaptability or versatility

If image data is fetched from multiple rows during rotation, then image rotation capability is achieved, but memory access efficiency deteriorates due to page changes

Engineering Contradiction:
Improveimage rotation capabilityVSAvoidmemory access efficiency
Core Design Contradiction:
Adaptability or versatilityVSProductivity

Solution Approach 1:

The patent applies segmentation by dividing the image data into alternating row segments that are stored in an interleaved manner. Instead of storing all pixels from one row contiguously, the memory is segmented to hold pixels from multiple rows in an alternating sequence. This segmentation allows the rotation operation to access pixels sequentially from a single memory page, maintaining high memory access efficiency while achieving full rotation capability.

Inventive Principle:
Principle #1Segmentation

3Ease of manufacture

If DRAM/SDRAM is used for storing image data, then cost is reduced compared to SRAM, but access speed decreases due to refresh requirements

Engineering Contradiction:
ImprovecostVSAvoidaccess speed
Core Design Contradiction:
Ease of manufactureVSSpeed

Solution Approach 1:

The patent applies parameter changes by modifying the memory access pattern and data organization rather than changing the fundamental memory technology. By reorganizing how pixels are stored in DRAM/SDRAM (interleaved row storage), the system achieves SRAM-like access speeds for rotation operations while maintaining the cost advantages of DRAM/SDRAM. This parameter change in data organization allows the system to overcome the inherent speed limitations of DRAM/SDRAM without switching to more expensive memory technology.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentUS7512287B2Method and apparatus for efficient image rotation
Publication Date: 2009.03.31 SEIKO EPSON CORP
  • US7512287B2 patent drawing
  • US7512287B2 patent drawing
  • US7512287B2 patent drawing

AI summary

The invention is generally directed to a method and apparatus for efficient rotation of an image. The image is preferably defined by an image matrix in which each datum of the image corresponds to a particular image row and a particular image column of the image matrix. A preferred method includes mapping each of the datum into a memory matrix organized into a plurality of memory rows and a plurality of memory columns. The mapping preferably includes mapping into one of the memory rows: (a) at least two data, both of which correspond to one of the image rows, and (b) at least two image vector portions, a first one of the image vector portions including at least two data each corresponding to one of the image columns, and a second one of the image vector portions including at least two other data each corresponding to another of the image columns. In addition, preferably the at least two data of the first image vector portion and the at least two other data of the second image vector portion each respectively correspond to the same image rows.