Salient area detection method based on non-end-to-end deep learning network

Abstract

The invention discloses a salient area detection method based on a non-end-to-end deep learning network. The method comprises the following steps: 1 generating a training sample; 2 deep network construction: constructing a four-segment convolutional neural network; 3 carrying out deep network training; 4 saliency detection: importing a test sample into the network model obtained through training in the step 3, obtaining the probability that an area block belongs to a saliency area through a softmax classifier, namely the saliency value of a superpixel corresponding to the area block, and enabling the saliency values of all the superpixels in the test sample to form a color saliency map and a texture saliency map of the test sample; and 5 saliency map fusion: fusing the color saliency map and the texture saliency map according to an adaptive weighting mode to obtain a final saliency map. According to the method, the deep learning network is adopted for image feature extraction, detection of salient areas and objects in a complex image scene can be achieved, and the problem that the detection rate is low in the complex image scene is solved.

Description

technical field [0001] The invention relates to the technical fields of image processing and computer vision, in particular to a salient region detection method based on a non-end-to-end deep learning network. Background technique [0002] Image saliency can be described as the ability of constituent elements in an image to attract human visual attention. Saliency detection can locate important areas in a scene without any prior knowledge, helping to quickly detect target areas in images or videos, It can be used in scenarios such as autonomous driving, intelligent security, and social networking. [0003] The term "saliency" is related to the foreground / background contrast, based on which saliency detection methods directly employing contrast measures have been generated. Cheng et al [Cheng M, Mitra N J, Huang X, et al.Global contrast basedsalient region detection[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2015,37(3):569-582] proposed a global contra...

AI Technical Summary

Problems solved by technology

[0006] The present invention provides a saliency region detection method based on a non-end-to-end deep learning network to solve the low detection rate in complex image scenes of the saliency detection method directly using contrast measurement in the prior art, and the end-to-end problem Missing Global Contrast in Deep Learning Saliency Detection Methods

Images