Image Signal Processor Filter Coefficient Sharing for Memory Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing demand for high-resolution images with higher quality has led to a need for reducing the number of filters used in de-mosaic processes, which in turn requires minimizing memory usage for filter coefficients in image processing devices.
Innovation Solution
An image processing device and method that utilize an image sensor with unit groups of pixels and color filters, along with an image signal processor that calculates a hash of pixel properties based on brightness or slope information, converts pixel indexes and properties, and adjusts filter coefficients accordingly to perform filtering, thereby reducing the number of filters needed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If the number of filters is increased to achieve high-resolution images with higher quality, then image quality is improved, but memory usage for storing filter coefficients increases
Solution Approach 1:
The patent applies universality by making filters in the conversion group share the same filter coefficients as their corresponding reference group filters. Instead of having separate filters for each group, the conversion group filters are configured to use the reference group's coefficients, allowing one set of filters to serve multiple purposes and reducing overall memory requirements while maintaining image quality.
Solution Approach 2:
The patent uses copying by having the conversion group filters copy the filter coefficients from the reference group filters. The filter coefficients of the reference group are copied and applied to the conversion group, eliminating the need to store separate coefficient sets for each group and reducing memory usage while preserving filtering effectiveness.
2Quantity of substance
If the number of filters is reduced to decrease memory usage, then memory footprint is reduced, but image quality may deteriorate
Solution Approach 1:
The patent applies universality by making filters in the conversion group share the same filter coefficients as their corresponding reference group filters. Instead of having separate filters for each group, the conversion group filters are configured to use the reference group's coefficients, allowing one set of filters to serve multiple purposes and reducing overall memory requirements while maintaining image quality.
Solution Approach 2:
The patent uses copying by having the conversion group filters copy the filter coefficients from the reference group filters. The filter coefficients of the reference group are copied and applied to the conversion group, eliminating the need to store separate coefficient sets for each group and reducing memory usage while preserving filtering effectiveness.
Data Source
AI summary
An image processing device comprises an image sensor including unit groups comprising a reference group and a conversion group each including pixels and color filters, and is configured to generate image data using the pixels, and an image signal processor which is configured to calculate a hash of the image data, and to select a filter corresponding to the hash to perform filtering, wherein the image signal processor includes a pixel index converter which is configured to convert indexes of the pixels based on a positional relationship between the pixels, a pixel property converter which is configured to convert the properties of the pixels and to convert the hash before conversion, and a filter coefficient converter which is configured to convert filter coefficients of the pixels based on the respective converted pixel index and the respective converted pixel property.