Infrared image salient region detection method based on hybrid algorithm

Abstract

The invention discloses an infrared image salient region detection method based on a hybrid algorithm. The method includes the following steps: firstly, carrying out local sparse representation on an original infrared image to get an initial salient map based on local sparse representation; secondly, putting forward an infrared image foreground prediction algorithm integrating the 2D maximum entropy / minimum cross entropy and the maximum standard deviation to process the original infrared image to get a target-highlighted and background-suppressed foreground prediction map; thirdly, subtracting the foreground prediction map from the original image to get a background map of the original infrared image; and finally, subtracting the background map from the initial salient map to get a final salient region detection result. Compared with the initial salient map, the salient region in the result has a clearer contour, and the non-salient region is fully suppressed. The salient region of an infrared image is detected by mixing the local sparse representation theory, the information entropy and the maximum standard deviation, and a more accurate result can be obtained.

Description

technical field [0001] The invention relates to a method for detecting a salient region of an infrared image based on a hybrid algorithm, and belongs to the technical field of image processing. Background technique [0002] With the rapid development of image processing technology, image visual saliency detection has become a topic of much concern. Saliency detection is widely used in various fields of image processing, such as image segmentation, image search, object detection and recognition, etc. The saliency detection of the image is to let the computer quickly identify the saliency area that contains useful information in the infrared image and is the most concerned by human vision. The result of saliency is the ability to obtain saliency maps of highlighted saliency regions that are complete, clear, and free from background interference. [0003] At present, my country's relevant patents in this field include: an image visual saliency detection method based on moving...

AI Technical Summary

Problems solved by technology

Although this detection method can highlight salient objects and suppress the image background well, there are still shortcomings that the salient objects obtained by this method are uneven and cannot preserve the complete boundary of the target.

[0004] Image visual saliency calculation method based on color histogram and global contrast (public number: CN102129693A), this method considers global contrast and spatial coherence at the same time, but because only the color histogram and spatial relationship are used, the obtained saliency map is Based on segmented regions instead of pixels, the final saliency map is rough, and the detection results are not ideal when the background is complex

[0005] To sum up, there are quite a lot of methods for image saliency detection at present, but there are still deficiencies. The algorithm is not mature enough. The shortcomings are reflected in the uneven highlighting of the target, the easy loss of boundary information, and the difficulty in facing complex backgrounds. Remove background noise, there is still a lot of background interference

Images