Terminal, image processing method and device thereof

Abstract

The invention provides a terminal, an image processing method and a device thereof. The image processing method comprises the steps of obtaining a texture consistent region based on the channel parameter of a first image, determining an edge area in the first image, filling the texture consistent region, and filling the edge area to separate the texture consistent region. According to the technical scheme of the invention, the process of image special effect processing is simple and the self adaptability is strong, the amount of calculation is small and the speed is fast.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to a terminal and an image processing method and device. Background technique [0002] With the development of digital media technology, special effects processing of images has received extensive attention and research. Taking the cartoonization of video or image as an example, it is also called Image stylization or Image abstraction, which belongs to the research field of Non-photorealistic rendering. Compared with photorealistic rendering , NPR is more focused on the personality of the image. Cartoon images are used in many fields, such as video conferencing, online games, etc. In addition, using virtual cartoon avatars instead of real faces can improve the data transmission speed of the network on the one hand, and create a relaxed and harmonious atmosphere on the other hand. atmosphere. [0003] At present, there are two methods for obtaining cartoon images:...

AI Technical Summary

Problems solved by technology

However, when the geometric model is used to cartoonize the target area, accurate geometric features need to be extracted, which leads to a large amount of calculation and high quality requirements for the photos

In addition, since the reference overall structure of the shape of the target area is defined in advance, for the actually captured photos, the geometric features of the target area may not be unified into one of the reference overall structures of the target area due to different shooting conditions. In the framework, for example: the reference overall structure of the front face is defined in advance, but in actual shooting, only the side of the face may be captured, so only one eye of the person is shown in the photo, and the actual extracted geometric features cannot correspond to In the reference overall structure of the front face defined in advance, it cannot be cartoonized

Another example is that during the shooting process, if the light is too strong, people will squint their eyes. At this time, the round eyes will become oval eyes. The oval eyes cannot correspond to the reference overall structure of the pre-defined face area, resulting in the failure of cartoonization.

For another example, when the eyes are closed, the eyes will be blocked by the eyelids, so that the eyes cannot be extracted, and the face cannot be cartoonized according to the reference overall structure of the face area.

Therefore, the adaptability of the geometric model method is poor

Images