18055603. APPARATUS FOR PERFORMING FILTER PROCESSING USING CONVOLUTION OPERATION, METHOD OF PERFORMING FILTER PROCESSING, AND MEDIUM simplified abstract (CANON KABUSHIKI KAISHA)
Contents
- 1 APPARATUS FOR PERFORMING FILTER PROCESSING USING CONVOLUTION OPERATION, METHOD OF PERFORMING FILTER PROCESSING, AND MEDIUM
- 1.1 Organization Name
- 1.2 Inventor(s)
- 1.3 APPARATUS FOR PERFORMING FILTER PROCESSING USING CONVOLUTION OPERATION, METHOD OF PERFORMING FILTER PROCESSING, AND MEDIUM - A simplified explanation of the abstract
- 1.4 Simplified Explanation
- 1.5 Potential Applications
- 1.6 Problems Solved
- 1.7 Benefits
- 1.8 Original Abstract Submitted
APPARATUS FOR PERFORMING FILTER PROCESSING USING CONVOLUTION OPERATION, METHOD OF PERFORMING FILTER PROCESSING, AND MEDIUM
Organization Name
Inventor(s)
APPARATUS FOR PERFORMING FILTER PROCESSING USING CONVOLUTION OPERATION, METHOD OF PERFORMING FILTER PROCESSING, AND MEDIUM - A simplified explanation of the abstract
This abstract first appeared for US patent application 18055603 titled 'APPARATUS FOR PERFORMING FILTER PROCESSING USING CONVOLUTION OPERATION, METHOD OF PERFORMING FILTER PROCESSING, AND MEDIUM
Simplified Explanation
The patent application describes an apparatus for performing filter processing on a data array in a processing target block of a predetermined size. The apparatus includes a data memory, a coefficient memory, a controller, and a processor.
- The data memory holds the data array in the processing target block.
- The coefficient memory holds weight coefficients of a filter used for the filter processing.
- The controller determines whether data in a reference region in the processing target block are zero values.
- The processor generates a convolution operation result of the weight coefficients and data at multiple positions in the processing target block.
- The controller controls whether to perform multiply-accumulate operations of the data and the weight coefficients based on the result of the determination.
Potential Applications
- Image and video processing: The apparatus can be used for various image and video processing tasks such as noise reduction, edge detection, and image enhancement.
- Signal processing: It can be applied to filter and analyze signals in various domains such as audio, radar, and telecommunications.
- Data analysis: The apparatus can be used for filtering and processing large datasets in fields like data mining, machine learning, and scientific research.
Problems Solved
- Efficient processing: The apparatus optimizes the filter processing by determining whether certain data values are zero, reducing the number of multiply-accumulate operations required.
- Memory management: The data memory and coefficient memory efficiently store and access the data array and weight coefficients, respectively, for the filter processing.
- Real-time processing: The apparatus enables fast and real-time processing of data arrays, making it suitable for applications that require quick analysis and response.
Benefits
- Improved performance: By reducing the number of multiply-accumulate operations, the apparatus enhances the efficiency and speed of filter processing.
- Memory efficiency: The data memory and coefficient memory efficiently utilize memory resources, allowing for larger data arrays and more complex filters to be processed.
- Versatility: The apparatus can be applied to various domains and applications, providing flexibility in implementing different filter processing tasks.
Original Abstract Submitted
An apparatus for performing filter processing on a data array in a processing target block of a predetermined size is provided. A data memory holds the data array in the processing target block. A coefficient memory holds weight coefficients of a filter used for the filter processing. A controller determines, in a determination, whether data in a reference region in the processing target block, set in correspondence with the processing target block, are zero values. A processor generates a convolution operation result of the weight coefficients and data at a plurality of positions in the processing target block. The controller controls, based on a result of the determination, whether to perform at least some of multiply-accumulate operations of the data and the weight coefficients when the processor generates the convolution operation result.
