A Simple and Fast Method for Positioning Horizontal Headlines and Subtitles in News Images

Abstract

The invention relates to a news image horizontal headline caption simply and rapidly positioning method. The news image horizontal headline caption simply and rapidly positioning method comprises collecting and inputting an image into a computer; capturing the lower area of the entire image as an area to be processed; calculating the gradient of every point of the area to be processed, and obtaining the edge graph of the area to be processed through threshold values; according to the number of edge points contained in every row in the edge graph, obtaining the positions of the top boundary and the bottom boundary; for the obtained positions of the top boundary and the bottom boundary, obtaining a corresponding target area on the basis of the maximum feature score; by means of the gradient information value of the target area, obtaining a first headline caption area; obtaining the others headline caption areas according to the position difference and the height ratio of a left boundary; outputting the boundary position information of the obtained headline caption areas. The news image horizontal headline caption simply and rapidly positioning method can help simply and rapidly position a horizontal headline caption in a news image.

Description

technical field [0001] The invention relates to the field of automatic detection of image features in computer vision, in particular to a simple and fast positioning method for horizontal titles and subtitles in news images. Background technique [0002] The location of captions and subtitles in news images has important applications in content-based news video or image retrieval, indexing and understanding. For a long time, the localization methods of horizontal titles and subtitles in news images can be divided into three categories: 1) projection-based methods, which are easily affected by punctuation marks, resulting in the truncated title subtitle area; 2) color-based methods, which It is easily disturbed by the background color, and the computational complexity is relatively high; 3) The method based on machine learning requires a large number of samples for training, resulting in a complex algorithm and low efficiency. Contents of the invention [0003] The inventi...

AI Technical Summary

Problems solved by technology

For a long time, the localization methods of horizontal titles and subtitles in news images can be divided into three categories: 1) projection-based methods, which are easily affected by punctuation marks, resulting in the truncated title subtitle area; 2) color-based methods, which It is easily disturbed by the background color, and the computational complexity is relatively high; 3) The method based on machine learning requires a large number of samples for training, resulting in complex algorithms and low efficiency

Images