Method for automatic generation of ternary map and foreground extraction method using the ternary map

Abstract

The invention relates to an automatic generation method for a ternary diagram, and the method comprises the steps: S1, obtaining an input image, and continuously scanning the image through windows; S2, extracting the image edge in each window through an edge detection algorithm, and obtaining an edge image of the input image; S3, obtaining an incomplete mark input image of an image segmentation algorithm according to the edge image; S4, obtaining a segmentation result image through the image segmentation algorithm and the incomplete mark input image; S5, obtaining the ternary diagram accordingto the segmentation result image. The method which does not need the interactive input of a user can achieve the full-automatic extraction of the foreground of a digital image without a special device. Moreover, the algorithm is better in real-time performances and extraction precision.

Description

Technical field [0001] The invention relates to an image processing method, in particular to a method for automatically generating a ternary image and a foreground extraction method using the ternary image. Background technique [0002] Digital image matting refers to a technology that naturally extracts the foreground of any shape from an image or video sequence. According to a small amount of user interaction information, the information needed in the image is automatically and accurately solved, that is, automatic matting, and the fully automatic matting that does not require user input. Automatic matting is a key technology in the field of image processing, which can greatly reduce the workload of manual matting and significantly improve work efficiency. [0003] At present, the more mature and widely used automatic matting algorithm is the Grab Cut matting algorithm of Microsoft Research. We can use this algorithm for matting in Microsoft Office PowerPoint. The matting functi...

AI Technical Summary

Problems solved by technology

Strictly speaking, the above two methods are semi-automatic image matting, requiring user interaction to output the matting results, and the improvement of work efficiency is limited.

However, automatic matting mostly relies on special equipment, such as depth images, binocular images, mirror reflections, etc., to solve the depth information of the foreground target and then matting. This method is difficult to popularize due to the special requirements of imaging.

The cutout cloud service developed by Alibaba Cloud Computing Co., Ltd. can only process monochrome background images, and the edges of the foreground are jagged. The application scenarios are limited and the effect is not good.

And some matting algorithms with better effects, such as Closed-Form Matting, require the input of the original image and the corresponding user graffiti marks (use a brush to randomly smear the determined foreground and background), while algorithms such as SharedMatting require the user to provide the original image and its corresponding trimap (a ternary image of the same size as the original image, with white for the determined foreground part, black for the determined background part, and the gray part for the cutout area to be solved), usually the pictures do not have the corresponding trimap , and how to get trimap brings a certain burden to users, so these excellent algorithms are difficult to popularize and apply

Images