Character image automatic segmentation method based on deep learning and information data processing terminal

Abstract

The invention belongs to the technical field of image processing, and discloses a character image automatic segmentation method based on deep learning and an information data processing terminal. Themethod comprises: collecting character pictures to form a training data set; constructing a deep neural network model of first-level image semantic segmentation; inputting the collected training dataset into a first-stage deep neural network to generate a trimap; constructing a second-level deep neural network model; inputting the collected training data set and the obtained trimap into a second-level deep neural network to generate a segmented character mask image; and synthesizing the character mask image and the figure original image to obtain a segmented character image. According to thecharacteristics of the character image, the character and the image background of the character image are automatically segmented, characters in the image are automatically screened, and the characters from the background picture are separated in combination with character features. The method can be used for automatic character matting, and can also be used for character photo background replacement and background processing through background blurring.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to a method for automatically segmenting a person image based on deep learning and an information data processing terminal. Background technique [0002] At present, the closest existing technology: person image segmentation refers to the separation of the foreground and background of the person photo, the goal is to classify each pixel of the input picture: foreground and background, and obtain the classification map at the pixel level. Image matting has a history of more than 30 years. At present, many mature algorithms still need to rely on manual participation of users, which is inefficient and lacks automation. Many image segmentation algorithms appeared in the early days, such as threshold segmentation algorithm, edge-based segmentation algorithm, region expansion algorithm, watershed algorithm, etc. segmentation results. Researchers try to add user inte...

AI Technical Summary

Problems solved by technology

Although many deep learning-based methods have emerged in image semantic segmentation in recent years, such as FCN and its derivative methods such as SegNet and DeepLab, these methods aim to achieve general image semantic segmentation, but have not achieved high-precision character image segmentation. The main reason for not achieving high precision is that these methods cannot finely segment the edge of the object, and the error caused by the upsampling part of the learning network has not been eliminated.

[0003] To sum up, the problems existing in the existing technology are: the existing method based on deep learning image semantic segmentation has the problem of low segmentation accuracy of character images; relying on manual interactive input, it is difficult to achieve automatic image matting, and a large amount of image processing is required The time efficiency is low; there is a problem of blurred boundaries in the automatic figure cutout

[0004] Difficulty in solving the above technical problems: One of the difficulties in automatic segmentation of person images lies in the accuracy of person matting. The backgrounds of character images are rich and diverse, and it is very difficult to realize automatic matting under complex and changeable backgrounds, especially when the background and character colors are different. When they are similar, it is difficult to rely on ordinary image processing methods for segmentation; the second difficulty lies in the edge processing of the characters, especially where the characters' hair and background are mixed, it is difficult to achieve accurate segmentation through the existing general semantic segmentation methods, and it is difficult to achieve Automatically and accurately segment hair strands in human images

[0005] The significance of solving the above technical problems: At present, there are still a lot of human intervention in the field of figure matting. For example, when the background needs to be replaced or blurred, the designer needs to spend a lot of time to extract the characters from the photos of the characters. When processing a large number of pictures particularly inefficient

Images