Applicability prediction and performance blind evaluation method of visual saliency detection algorithm

Abstract

The invention discloses a method for applicability prediction and performance blind evaluation of a visual saliency detection algorithm, which relates to the technical field of identification methods, and includes an algorithm applicability analysis module and a saliency detection performance blind evaluation module. Under the condition of the original image, predict and screen the candidate saliency method most suitable for the image through the applicability prediction network, and then input the saliency map generated by the candidate saliency method into the saliency detection performance blind evaluation module, and finally sort, get Best saliency map. When performing blind evaluation of saliency detection performance, it is usually necessary to generate the saliency map of each algorithm in advance, and then input it into the quality evaluation network for sorting, and select the saliency map with the highest score, but generating the saliency map of each algorithm takes most of the time, reducing The efficiency of quality evaluation is improved, and the visual saliency detection algorithm applicability prediction and performance blind evaluation method solves this technical problem, and also has the advantages of comprehensiveness and high precision.

Description

technical field [0001] The invention relates to the technical field of identification methods, in particular to a method for applicability prediction and blind performance evaluation of a visual saliency detection algorithm. Background technique [0002] Visual saliency detection aims to simulate human visual characteristics by designing intelligent algorithms, automatically detect salient areas in digital images that are easily noticed by observers, and mark and display them in the form of heat maps. Due to the important role of visual saliency detection in various advanced visual perception algorithms, this problem has received extensive attention and research in recent years. Visual saliency detection can be applied to many advanced computer vision problems, such as image compression, image stitching, image segmentation, etc. Given the image to be detected, the visual saliency detection algorithm outputs a pixel-level saliency prediction map. Taking the output gray salie...

AI Technical Summary

Problems solved by technology

In addition, the visual patterns in natural images vary greatly, and different visual saliency detection methods have different design ideas, and various methods are often suitable for different types of images.

Images