Image Signal Processor Depth Map Generation Speed Accuracy

Solution Overview

Problem

Existing image signal processors face challenges in balancing the speed and accuracy of depth map image generation, particularly when reducing image resolution for faster processing, which often compromises depth map accuracy.

Innovation Solution

The proposed image signal processor includes a depth map generator that downsizes input images to target resolution and generates depth maps based on disparity values, and a noise remover that performs guide filtering on the depth map using the original input image as a guide to enhance accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If image resolution is decreased to increase depth map generation speed, then generation speed is improved, but depth map accuracy deteriorates

Engineering Contradiction:

Improvedepth map generation speedVSAvoiddepth map accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The processing is divided into two stages: first, downsizing the left and right images to a lower resolution for rapid depth map generation; second, applying guide filtering using the original high-resolution image to refine and enhance the depth map accuracy. This segmentation allows the system to achieve both speed (from the first stage) and accuracy (from the second stage).

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system performs preliminary downsizing of the input images before generating the depth map, which enables faster processing. The original high-resolution image is preserved and used later as a guide for filtering, ensuring that the preliminary action does not compromise final accuracy.

Inventive Principle:

Principle #10Preliminary action

2Measurement precision

If guide filtering is applied using the original input image, then depth map accuracy is enhanced, but processing time increases

Engineering Contradiction:

Improvedepth map accuracyVSAvoidprocessing time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The processing is divided into two stages: first, downsizing the left and right images to a lower resolution for rapid depth map generation; second, applying guide filtering using the original high-resolution image to refine and enhance the depth map accuracy. This segmentation allows the system to achieve both speed (from the first stage) and accuracy (from the second stage).

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The guide filtering operation uses the original high-resolution image as a guide, which provides more information than strictly necessary but ensures high accuracy. The filtering process selectively uses guide image data only where needed to refine the depth map, balancing the additional processing time with the significant accuracy improvement.

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS20250086818A1Image signal processor and depth map generation method

Publication Date: 2025.03.13 SK HYNIX INC

AI summary

The present technology relates to an image signal processor. According to the present technology, an image signal processor may include a depth map generator configured to generate, based on a target resolution of an output image, a downsized left image and a downsized right image based on an input image received from an external device, and generate a depth map image indicating depth values corresponding to pixels included in the output image based on the downsized left image and the downsized right image, and a noise remover configured to perform guide filtering on the depth map image using the input image as a guide image and generate the output image.