AI Image Outpainting via Segmented Stitching for Large Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
High-resolution image out-painting using AI biographic technology consumes significant system resources, making it challenging for devices like notebook computers to handle large-sized images.
Innovation Solution
An image out-painting method and system that cuts an original image into multiple segments, processes each segment individually, and then stitches them together to form a large-sized image, utilizing pipeline technology to avoid simultaneous resource occupation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If high-resolution image out-painting is performed using AI biographic technology, then image quality and resolution are improved, but system resource consumption increases significantly
Solution Approach 1:
The patent divides the original image into multiple smaller sub-images and processes them separately through the AI out-painting model. This segmentation allows the system to handle large images by breaking them into manageable chunks, reducing the memory and computational resources required at any given time while maintaining overall image quality.
Solution Approach 2:
The patent performs preliminary actions by first dividing the image into sub-images and preparing them for processing before applying the AI out-painting model. This pre-processing step enables efficient resource management by ensuring that only the necessary portions of the image are loaded into memory at each processing stage.
2Area of stationary object
If large-sized images are processed using AI out-painting, then image expansion capability is improved, but memory resource occupation increases
Solution Approach 1:
The patent applies segmentation by dividing large images into smaller sub-images that can be processed individually. This approach enables the system to expand large images beyond what would be possible with single-image processing, as each sub-image consumes less memory while the final result achieves the desired large size.
Solution Approach 2:
The patent transitions from processing a single large image in two dimensions to processing multiple smaller sub-images in a grid arrangement. This dimensional change allows the system to manage memory resources more effectively by distributing the processing load across multiple smaller units rather than attempting to load the entire large image at once.
3Device complexity
If the entire original image is processed at once, then processing simplicity is maintained, but calculation resource occupation increases
Solution Approach 1:
The patent divides the image processing task into segmented sub-tasks, where each sub-image is processed separately. While this introduces additional steps for division and reassembly, it dramatically reduces the calculation resources required at each processing moment, making the overall system more feasible for devices with limited computational power.
Data Source
AI summary
An image out-painting method and an image out-painting system are provided. The image out-painting method includes the following steps. An original image is provided. A ring block surrounding the original image is filled with a plurality of edge pixels of the original image to obtain a pre-processed image. The pre-processed image is cut into a plurality of to-be-processed images, each including a part of the original image and a part of the ring block. A plurality of masks are obtained according to the to-be-processed images. The contents of the ring block in the to-be-processed images are individually out-painted according to the masks to obtain a plurality of painted images. The painted images are stitched together to form an expanded image.