Detection of Lesions in Regularized Gastroscopy Images Based on Dual Attention Mechanism

摘要

A method for lesion detection based on image low-rank sparse decomposition is proposed. First, the algorithm performs initial processing on the input gastroscopy image, extracts the color and texture features of the image to obtain the initial image, and then divides the initial image into blocks, performs low-rank decomposition on each equally divided small block, and combines to obtain a locally decomposed image. The image is directly subjected to low-rank decomposition processing to obtain a globally decomposed image. Weighted fusion is performed on the two decomposed images to obtain the final decomposed image, and the interference content of the image is removed to obtain the detection result of the lesion area.