A dermatoscope image retrieval method based on end-to-end deep hashing

Abstract

The invention discloses a dermatoscope image retrieval method based on end-to-end depth hashing. The dermatoscope image retrieval method comprises the steps of 1, establishing a dermatoscope image database; 2, designing an end-to-end deep hash network model; 3, carrying out network training; 4, extracting a deep hash code, and constructing a retrieval database; And 5, retrieving the dermatoscope image. The method has the advantages that by designing a Res-DenseNet50 deep hashing structure, the fusion capability between high-level features and low-level features is improved, and the loss of information in the transmission process between layers is avoided. And the extracted high-level features have better separability, so that higher retrieval accuracy is achieved. The end-to-end deep hash-based retrieval method is realized. According to the method, the original image is directly learned, and the deep hash code corresponding to the input image can be directly obtained from the penultimate layer of the network, so that the dermatoscope image retrieval process is simplified, and the accumulative error between the front step and the rear step in the traditional retrieval process is avoided.

Description

technical field [0001] The invention belongs to the field of dermoscopic image processing, and in particular relates to an end-to-end depth hash-based dermoscopic image retrieval method. Background technique [0002] Various skin diseases in daily life endanger people's health, and dermoscopy diagnosis is a non-invasive microscopic image analysis technique for skin diseases. In the past, dermatologists mainly used dermatoscopy to observe the skin lesion area, and then relied on experience and subjective visual evaluation to make a diagnosis. This method of relying on human eye observation is likely to cause visual fatigue, and the diagnostic results are subjective and have poor repeatability. The dermoscopic image-assisted diagnosis technology can automatically extract skin lesion targets from dermoscopic images and identify the type of skin lesions, thereby assisting doctors to make a correct diagnosis. This method has the advantage of being objective and repeatable. Derm...

AI Technical Summary

Problems solved by technology

Since the current research on dermoscopic image retrieval is not very in-depth, the accuracy of existing dermoscopic image retrieval methods is not very high, and there is still a lot of room for improvement in the research on dermoscopic image retrieval

Images