A 3D Image Cropping Method Based on Example Learning and Enhanced Visual Quality

Abstract

The invention relates to a 3D image cropping method based on sample learning and enhanced visual quality, comprising: calculating the GIST feature and HIST feature of each image in a professional 3D stereoscopic image library; inputting the image to be cropped and the target size, and adopting a sampling method to obtain A set of candidate cropping windows; calculate the similarity between each candidate cropping window and the GIST feature and HIST feature of each image in the image library, and obtain the n images most similar to it, and combine them to form a learning sample set; use sample evaluation The composition and depth information distribution quality of each candidate cropping window, as well as information loss and compliance with stereoscopic image visual comfort rules, calculate the cropping window of the left view; through the learning sample, move the cropping window of the right view horizontally to obtain Similar depth distribution to obtain the final cropping result. This method is beneficial to obtain visually comfortable cropping results, and can be applied to image processing, computer vision and other fields.

Description

technical field [0001] The invention relates to the fields of image processing and computer vision, in particular to a 3D image cropping method based on example learning and visual quality enhancement. Background technique [0002] Image cropping is one of the most basic image processing operations. Since 2003, experts and scholars have continuously studied and proposed image cropping methods based on content awareness, that is, methods based on intelligent cropping. This type of method first crops according to the importance of the image information, initially obtains the cropped area with the same aspect ratio as the target screen, and then adapts the content of the cropped area to the size of the target screen using proportional scaling technology. Liu H. et al. determine the location information of important content in the image through eyetracking, and then cut out the content area containing the least important information in an interactive way. Liu F. et al. first d...

AI Technical Summary

Problems solved by technology

[0004] Unlike cropping for 2D images, cropping for 3D images is a challenging task because the quality of 3D images is highly subjective and the cropping window for 3D images may lead to violation of visual comfort rules for stereoscopic images

The difficulty of the 3D image cropping method is to ensure that the cropped results can be viewed comfortably through a stereoscopic display device

Existing automatic stereoscopic image cropping methods do not consider this problem

Images