Image Processing Rotation for SRAM-Saving Cross-Pattern Operations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image processing circuits require a large number of line buffers to store vertical pixels, leading to increased circuit cost due to the need for extensive static random-access memory (SRAM) when performing operations on pixels in a cross pattern.
Innovation Solution
An image processing method and apparatus that utilizes two stages of image rotations and row-direction operations to process pixels in a cross pattern, reducing the need for large line buffers by storing and processing images in a frame and operation storage circuit using dynamic random-access memory (DRAM) and static random-access memory (SRAM) configurations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If vertical pixels are accessed using SRAM in the form of line buffers, then image processing operation on cross pattern pixels can be performed, but circuit cost increases due to large amount of SRAM required
Solution Approach 1:
The patent transforms the problem from accessing vertical pixels in the original image coordinate system to accessing horizontal pixels in a rotated coordinate system. By rotating the image data by 90 degrees, the vertical direction becomes horizontal, allowing standard horizontal line buffer architecture to efficiently access what were originally vertical pixels, thereby reducing SRAM requirements while maintaining cross-pattern processing capability
Solution Approach 2:
The patent performs preliminary rotation of the image frame before the actual cross-pattern processing operation. The image is rotated 90 degrees in advance, and the processing algorithms are executed on this rotated frame. After processing, the result is rotated back by 270 degrees. This preliminary transformation allows the use of efficient horizontal line buffer access patterns throughout the processing pipeline
2Manufacturing precision
If a large number of pixels on vertical direction are stored, then better operation result is achieved, but more line buffers are required which increases circuit cost
Solution Approach 1:
The patent changes the dimensional orientation of pixel storage by rotating the entire image frame by 90 degrees. This transformation converts vertical pixel sequences into horizontal pixel sequences, allowing the use of standard horizontal line buffers to store and access the same pixel data that would otherwise require vertical line buffers, thereby reducing the quantity of buffers needed while maintaining operation quality
Solution Approach 2:
The patent changes the parameter of image orientation from 0 degrees (original orientation) to 90 degrees (rotated orientation). This parameter change fundamentally alters how pixels are arranged and accessed in memory, transforming vertical access patterns into horizontal access patterns that are more efficient for the available line buffer architecture
Data Source
AI summary
The present disclosure discloses an image processing method having storage resource saving mechanism including steps outlined below. An original image frame is stored in a frame storage circuit and is read column-by-column to an operation storage circuit. The original image frame is rotated along a rotation direction for 90 degrees to generate a first rotated image frame. A first row direction image operation is performed row-by-row to generate a first operated image frame to be stored in the operation storage circuit. The first operated image frame is rotated along the rotation direction for 270 degrees to generate a second rotated image frame. The second rotated image frame is stored in the frame storage circuit. The second rotated image frame is read column-by-column to the operation storage circuit. A second row direction image operation is performed row-by-row to generate a second operated image frame. The second operated image frame is outputted.