Image Processor Motion Search Speed via Internal Memory Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing image processor for motion search requires significant data transfer between the motion search unit and frame memory, leading to latency due to the use of SDRAM with low operational speed, resulting in elongated motion search times.
Innovation Solution
The image processor employs internal memories with higher access speeds than external memory to perform motion search, including an image production unit, search units, and a transfer unit to read and process images within these internal memories, reducing the time required for data access and achieving high-speed motion search.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If data is read from external memory (frame memory) for motion search, then the motion search can be performed with available reference images, but the access speed is slow causing latency and elongated search time
Solution Approach 1:
The patent divides the memory system into multiple segments: external frame memory for storing complete reference images, and multiple internal memory banks (first internal memory, second internal memory, third internal memory) for storing specific ranges or processed versions of images. This segmentation allows parallel access to different data portions, significantly improving access speed and reducing motion search time.
Solution Approach 2:
The patent introduces an additional dimension of memory hierarchy by adding internal memory banks between the external frame memory and the motion search unit. This creates a multi-level memory structure where data can be accessed from different levels depending on the search requirements, enabling faster access without losing the capability to access complete reference images from external memory.
2Productivity
If multiple images are stored and accessed from external memory for different search stages, then comprehensive motion search can be performed, but the data transfer volume increases causing more latency
Solution Approach 1:
The patent performs preliminary actions by pre-loading required image ranges into internal memory banks before the actual motion search begins. The image production unit reads and processes images in advance, storing results in internal memories. This preliminary preparation eliminates the need for repeated data transfer during the search process, significantly improving productivity while reducing data transfer time.
Solution Approach 2:
The internal memory banks serve as intermediary storage between the external frame memory and the motion search unit. Instead of directly transferring large volumes of data between external memory and the search unit, the intermediary internal memories buffer and pre-process the data, reducing the total data transfer volume and time while maintaining the capability to access multiple images for different search stages.
Data Source
AI summary
An image processor includes an LSRAM accessible with a higher speed than a frame memory and configured to hold a second image in a predetermined range of a first image, an image production unit configured to read an image in a predetermined range of the second image and produce a third image for rough search based on the read image, an MSRAM accessible with a higher speed than the frame memory and configured to hold the third image, a first search unit configured to read the third image and perform first motion search based on the third image, and a second search unit configured to read a fourth image in a predetermined range of the second image based on a search result by the first search unit and perform second motion search that is more detailed than the first motion search based on the fourth image.