Image Processing Apparatus Selective Neural Network Block Recovery

Solution Overview

Problem

Existing image processing technologies face limitations in reducing calculation loads when using neural networks for image recovery after decoding, particularly due to the high computational demands of neural network calculations in integrated devices.

Innovation Solution

An image processing apparatus and method that assesses the degree of degradation in encoded image blocks and selectively applies image recovery processing using a neural network only where necessary, reducing overall calculation requirements by switching between different neural network configurations based on block degradation levels.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If image recovery processing using a neural network is executed on all blocks in the decoded image, then image quality is improved, but the amount of calculations increases significantly

Engineering Contradiction:

Improveimage qualityVSAvoidamount of calculations

Core Design Contradiction:

Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent applies local quality by differentiating processing strategies based on block characteristics. Blocks are classified into different types (e.g., high-frequency blocks, low-frequency blocks, blocks with high degradation) and only certain blocks undergo neural network recovery processing. This selective application of image recovery processing to specific blocks rather than all blocks reduces the overall calculation load while maintaining image quality in the most critical areas.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The patent segments the decoded image into multiple blocks and evaluates each block individually for degradation level. By dividing the image processing task into block-level segments and applying recovery processing only where necessary, the system reduces total calculations while preserving image quality in degraded regions.

Inventive Principle:

Principle #1Segmentation

2Manufacturing precision

If image recovery processing using a neural network is executed on all blocks in the decoded image, then image quality is improved, but processing time increases

Engineering Contradiction:

Improveimage qualityVSAvoidprocessing time

Core Design Contradiction:

Manufacturing precisionVSLoss of time

Solution Approach 1:

The patent applies local quality by differentiating processing strategies based on block characteristics. Blocks are classified into different types (e.g., high-frequency blocks, low-frequency blocks, blocks with high degradation) and only certain blocks undergo neural network recovery processing. This selective application of image recovery processing to specific blocks rather than all blocks reduces the overall calculation load while maintaining image quality in the most critical areas.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The patent applies partial action by executing image recovery processing on only a subset of blocks that meet specific criteria (e.g., high degradation level, high-frequency content) rather than processing all blocks. This partial processing approach reduces total processing time while still achieving acceptable image quality by focusing computational resources on the most problematic areas.

Inventive Principle:

Principle #16Partial or excessive action

3Manufacturing precision

If neural network processing is applied to recover image quality after decoding, then image quality degradation is reduced, but resource consumption in integrated devices increases

Engineering Contradiction:

Improveimage qualityVSAvoidresource consumption

Core Design Contradiction:

Manufacturing precisionVSUse of energy by moving object

Solution Approach 1:

The patent applies local quality by differentiating processing strategies based on block characteristics. Blocks are classified into different types (e.g., high-frequency blocks, low-frequency blocks, blocks with high degradation) and only certain blocks undergo neural network recovery processing. This selective application of image recovery processing to specific blocks rather than all blocks reduces the overall calculation load while maintaining image quality in the most critical areas.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The patent applies partial action by executing image recovery processing on only a subset of blocks that meet specific criteria (e.g., high degradation level, high-frequency content) rather than processing all blocks. This partial processing approach reduces total processing time while still achieving acceptable image quality by focusing computational resources on the most problematic areas.

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS11823422B2Image processing apparatus, control method of the same, and storage medium

Publication Date: 2023.11.21 CANON KK

AI summary

An image processing apparatus obtains a degree of degradation for blocks in an encoded image obtained by encoding an input image, decodes the encoded image; and executes image recovery processing on a block in a decoded image obtained by the decoding. The image processing apparatus, switches, depending on the degree of degradation of the blocks in the encoded image, whether or not the image recovery processing using a first neural network is executed on each block in the decoded image.