An image multi-scale feature recognition method based on image saliency

Image recognition refers to the technology which processes, analyzes and understands images with computer so as to recognize various targets and objects of different patterns. To effectively combine image recognition and intelligent algorithm can enhance the efficiency of image feature analysis, improve the detection accuracy and guarantee real-time detection. In image feature recognition, the following problems exist: the description of accurate object features, object blockage, complex and changeable scenes. Whether these problems can be effectively solved has great significance in improving the stability and robustness of object recognition algorithm. This paper takes image salience as the fundamental framework, and makes in-depth study of the problems of effective object appearance description, multi-feature fusion and multi-feature adaptive combination. Then it proposes an image multi-scale feature recognition method based on image salience and it can better locate the saliency object in the image, and more evenly highlight the salient object and significantly suppress background noises. The experiment results prove that salient region detection algorithm can better stress the entire salient image. © 2021, North Atlantic University Union NAUN. All rights reserved.