Device and method for confirming object region in image

Abstract

The invention provides a device for confirming an object region in an image to be confirmed. The device comprises an image segmentation module, a classifier training module and an object confirming module, wherein the image segmentation module is used for performing layered segmentation on a thermograph corresponding to the image to be confirmed so as to obtain a plurality of layers of layered images; each layer of layered images is segmented into a plurality of super-pixels; the classifier training module comprises a super-pixel classifier or a plurality of super-pixel classifiers respectively corresponding to a layer of layered images or a plurality of layers of layered images and is used for confirming the confidence degree of the super-pixels positioned in the object region or a non-object region in the layers corresponding to the super-pixel classifiers by using the super-pixel classifiers; and the object confirming module is used for confirming the confidence degree of the super-pixels positioned in the object region or the non-object region in the layers corresponding to the super-pixel classifiers on the basis of the super-pixel classifiers, and is used for confirming the object region in the image to be confirmed.

Description

technical field [0001] The present application relates to a device and method for determining a target area in an image, and more specifically, relates to a device and method for automatically determining a target area in a pending image by performing hierarchical segmentation on the pending image. Background technique [0002] So far, early detection of tumors remains a solution to reduce cancer mortality. For the past few decades, radiography has been the main modality used to diagnose tumors. However, this modality may have a high rate of false positives due to its low specificity and necessitates many unnecessary biopsies that are expensive and uncomfortable for the patient. [0003] Ultrasound technology is a more convenient and safer detection method, and can achieve higher accuracy in the diagnosis process. But it is well known that the interpretation of ultrasound images is extremely dependent on an understanding of the principles of contrast-enhanced sonography. ...

AI Technical Summary

Problems solved by technology

[0005] (1) The quality of the ultrasound image used to detect the tumor area is poor, with low contrast and severe speckle noise;

[0006] (2) The shape and appearance of the tumor are quite different;

[0007] (3) The tumor boundary is unclear and irregular;

Images