Saliency image detection method for interactive cycle feature remodeling

Abstract

The invention discloses a saliency image detection method for interactive cycle feature remodeling, and the method comprises the steps: constructing a convolutional neural network at a training stage,wherein the convolutional neural network comprises an input layer, a coding part, a decoding part and an output layer, the coding part comprises a neural network block, and the decoding part comprises an information extraction block, a feature remodeling block, an information remodeling block, an expansion convolution block and a feature aggregation block; inputting the three channels of the RGBimage of the 3D image and a three-channel depth map obtained by processing the depth image into a convolutional neural network for training to obtain a saliency detection map; calculating a loss function value between the saliency detection image and the label image to obtain an optimal weight vector and an optimal bias term; inputting three channels of an RGB image of a to-be-detected 3D image and a three-channel depth map corresponding to the depth image into the convolutional neural network training model in a test stage, and performing prediction by using the optimal weight vector and theoptimal bias term to obtain a saliency prediction image. The method has the advantages of clear saliency detection result and high detection precision.

Description

technical field [0001] The invention relates to a salient image detection technology of deep learning, in particular to a salient image detection method of interactive loop feature remodeling. Background technique [0002] With the rapid development of artificial intelligence in the computer field, image saliency detection has become a research field that has attracted more and more attention. Salient Object Detection (SOD) aims to distinguish the most visually distinctive objects from the input image, and over the past few decades, hundreds of traditional methods have been developed to solve the task of salient object detection , which is an efficient preprocessing step in many image processing and computer vision tasks, such as object segmentation and tracking, video compression, image editing, texture smoothing, etc. Recent work utilizes convolutional neural networks (CNNs) to learn deep features for detecting salient objects. These convolutional neural network models ad...

AI Technical Summary

Problems solved by technology

Existing saliency detection methods using the encoder-decoder structure of convolutional neural networks are quite effective, however, there are still challenges in accuracy

For example: features at different semantic levels and resolutions have different distribution characteristics, high-level features have rich semantic information, but lack accurate location information; low-level features have rich details, but are full of background noise, resulting in the fusion of high-level features The detection accuracy of the method with low-level features is still not very ideal

Images